fix(ef): RG 112(d) excludes a same-BVM commemoration (fix round 1, F1)

The BVM Saturday Office is itself "de B. Maria Virg."; RG 112(d) (Caput
XVI, "De Commemorationibus") excludes another commemoration whose own
oration invokes the SAME BVM's intercession. our-lady-of-mt-carmel (16
July, Commemoration_only) was wrongly admitted as an ordinary
commemoration alongside the Office on every 16-July-Saturday -- its own
collect explicitly invokes "her" (eius) patronage, and both photographic
scans print a rubric anticipating exactly this collision (an either/or
of Mass texts, not a joint commemoration).

Precedence_ef gains marian_slugs (a closed, hand-verified list of
sanctoral slugs that are themselves feasts of the BVM) and is_bvm_office
(true for subject = Bvm OR a marian_slugs member -- the disjunction is
load-bearing: the Office is identified only by subject, since its own
slug is deliberately reused from the ordinary ferial fallback; Mt Carmel
is identified only by the list, since its own subject tag is Saint, not
Bvm). disposition gains a new branch, checked before the pre-existing
Commemoration_only catch-all, since the one live witness is
Commemoration_only and that branch's own "always Commemorate" had no
exception for this shape.

Checked exhaustively against every Marian-named sanctoral entry: only
Commemoration_only-status entries can ever reach this branch live (a
Feast-status one always wins outright instead of losing). Of the two
such entries in the data, only Mt Carmel is reachable --
our-lady-of-ransom (24 September) is provably unreachable: 24 September
falling on a Saturday forces 1 September to a Thursday, which by
construction makes 24 September the September Ember Saturday, Class2,
every time, not a sampled coincidence.

Re-measured against the full pre-change binary: the original 75,853-day
blast radius is unchanged in total, splitting into 74,633 colour-only +
1,220 colour+comms (16-July Saturdays). Mutation-tested: exactly 3 new
tests redden, nothing else.

5 files changed, 324 insertions, 27 deletions

diff --git a/lib/rites/rite_ef/precedence_ef.ml b/lib/rites/rite_ef/precedence_ef.ml
index 5a2d307..302df98 100644
--- a/lib/rites/rite_ef/precedence_ef.ml
+++ b/lib/rites/rite_ef/precedence_ef.ml
@@ -364,14 +364,21 @@ let band (ctx : Vocab_ef.season Precedence.context) (c : Vocab_ef.rank Precedenc
   else if is_temporal && rank = Class3 && season = Advent then 250
     (* 26: III-class vigils (St Lawrence). *)
   else if rank = Class3 && is_vigil then 260
-    (* 27: Office of the BVM on Saturday -- every otherwise-unoccupied IV-class
-       Saturday, per the historical default that fills it; ordinary Mass
-       propers still make Rogation Mon/Tue/Wed proper without changing the
-       Office (RG 88, see temporal_ef.ml's [temporal]), so those never carry
-       this entry unless they happen to fall on the Saturday itself. Excludes
-       vigils for the same reason 11-13/14/16/19/20/23/24 do: RG 91 has no
-       IV-class vigil at all (RG 91's own vigil list, §4 "Vigils",
-       stops at III class), so one would be an anomaly, not this entry. *)
+    (* 27: Office of the BVM on Saturday -- RG 78 (Caput IX, "De sancta Maria
+       in sabbato"), both photographic scans and the electronic
+       transcription, word for word (docs/research/rules-register.md §4/§6.4;
+       Rite_ef.Temporal_ef's own [bvm_saturday_names] citation has the full
+       argument): "78. In sabbatis, in quibus occurrit Officium de feria IV
+       classis, fit de sancta Maria in sabbato" -- on Saturdays on which the
+       Office of a IV-class feria occurs, [the Office] is made of Holy Mary
+       on Saturday instead. Every otherwise-unoccupied IV-class Saturday
+       reaches this branch; ordinary Mass propers still make Rogation
+       Mon/Tue/Wed proper without changing the Office (RG 88, see
+       temporal_ef.ml's [temporal]), so those never carry this entry unless
+       they happen to fall on the Saturday itself. Excludes vigils for the
+       same reason 11-13/14/16/19/20/23/24 do: RG 91 has no IV-class vigil at
+       all (RG 91's own vigil list, §4 "Vigils", stops at III class), so one
+       would be an anomaly, not this entry. *)
   else if is_temporal && (not is_vigil) && rank = Class4 && weekday = Date.Sat then 270
     (* 28: IV-class ferias -- the unqualified catch-all (temporal_ef.ml's own
        comment on [ferial_rank] cites the same primary text, "Feriae IV
@@ -380,6 +387,148 @@ let band (ctx : Vocab_ef.season Precedence.context) (c : Vocab_ef.rank Precedenc
   else if (not is_vigil) && rank = Class4 then 280
   else unclassified
 
+(* RG 112(d) (Caput XVI, "De Commemorationibus"), the BVM half, fix round 1
+   of the ef-bvm-saturday task (coordinator finding F1) -- both photographic
+   scans, word for word (docs/research/rules-register.md §4's own RG 112
+   entry already carries this sub-clause, unbuilt until this fix): *"item,
+   Officium, Missa aut commemoratio de B. Maria Virg. aut de aliquo Sancto
+   vel Beato excludit aliam commemorationem aut orationem in qua eiusdem B.
+   Mariae Virg., vel Sancti aut Beati intercessio imploretur: quod tamen non
+   valet de oratione dominicae vel feriae, in qua fit invocatio eiusdem
+   Sancti."* -- the Office, Mass or commemoration OF the Blessed Virgin Mary
+   (or of some Saint or Blessed) excludes ANOTHER commemoration or oration
+   in which the intercession OF THE SAME BVM (or Saint or Blessed) is
+   invoked -- which however does not hold of the oration of a Sunday or
+   feria, in which invocation of the same Saint occurs.
+
+   Found live by the fix-round review: the BVM Saturday Office (RG 91 entry
+   27, [temporal_ef.ml]'s own [bvm_saturday_names]) is itself "de B. Maria
+   Virg." -- so when it is observed, RG 112(d) excludes any OTHER admitted
+   candidate whose own oration invokes the SAME BVM's intercession, not
+   merely commemorates her in passing. Real live witness: 16 July,
+   "our-lady-of-mt-carmel" (data/ef/sanctoral.sexp, Class3,
+   [Commemoration_only]) falls on a Saturday 8 times in the 2005-2050
+   fixture alone; its own collect, both photographic scans, word for word
+   (Caput "Die 16 iulii, Beatae Mariae Virginis de Monte Carmelo,
+   Commemoratio"): *"...concede propitius; ut, cuius hodie Commemorationem
+   solemni celebramus officio, EIUS muniti praesidiis, ad gaudia sempiterna
+   pervenire mereamur"* -- "her" ("eius") -- grant, we beseech Thee, that we,
+   fortified by HER patronage, whose Commemoration we solemnly celebrate
+   today, may be permitted to arrive at everlasting joys -- an explicit
+   invocation of the SAME BVM's own intercession/patronage, precisely RG
+   112(d)'s own trigger.
+
+   The exception clause ("non valet de oratione dominicae vel feriae") does
+   NOT rescue this: RG 91 lists entry 27 as its own table row, separate from
+   entry 28's "feriae IV classis" -- this Office is neither a Sunday nor,
+   once RG 78 has substituted it in, a plain "feria" in Caput IV's own
+   sense (RG 21: "Nomine feriae intelleguntur singuli dies hebdomadae" --
+   an ordinary weekday's OWN office, which RG 78 replaces, not merely
+   supplements).
+
+   ALSO found on the very same page, both photographic scans, word for
+   word, confirming the collision is real and rubric-anticipated (not
+   merely this codebase's own inference): *"Si Commemoratio B. Mariae
+   Virg. de Monte Carmelo venerit in sabbato, Missa dici potest aut de
+   sancta Maria in sabbato, aut propria de Commemoratione B. Mariae Virg.
+   de Monte Carmelo."* -- if the Commemoration of the BVM of Mount Carmel
+   falls on a Saturday, the Mass MAY be said EITHER of Holy Mary on
+   Saturday OR properly of the Commemoration of the BVM of Mount Carmel --
+   an explicit EITHER/OR between two MASS TEXTS, not an instruction to
+   commemorate one in the other's Office. Read together with RG 112(d):
+   the OFFICE question (is Mt Carmel commemorated at all) and the MASS
+   question (which of the two propers is said that day) are two different
+   questions -- RG 112(d) answers the first (excluded); this rubric answers
+   the second, and is a Mass-propers selection detail of exactly the same
+   kind as the I-V numbered cycle (RG 309(a)) [bvm_saturday_names]'s own
+   citation already puts out of scope for the SAME reason (colitur computes
+   no citations/readings at all yet, Plan 4) -- not modelled here either.
+
+   [marian_slugs] is a CLOSED, HAND-VERIFIED list, not a claim to have read
+   every one of these entries' own Latin orations on the scan (only
+   Mt Carmel's, quoted above, was actually verified against the primary
+   text for this fix) -- built from data/ef/sanctoral.sexp's own [names.en]
+   field, restricted to entries that are themselves a feast/commemoration
+   OF the Blessed Virgin Mary in her own right (never merely a feast of
+   someone else that happens to mention her -- "St. Anne, Mother of the
+   Blessed Virgin", "St. Joseph, Spouse of the Bl. Virgin Mary", "St.
+   Anthony Mary Claret" and similarly-named entries are deliberately
+   EXCLUDED, checked one by one). This is the SAME modelling simplification
+   {!disposition}'s own pre-existing RG 112(a) branch already makes for
+   "eiusdem Divinae Personae" (read as simply "both [subject = Lord]",
+   that branch's own comment: "safe at today's data's own granularity") --
+   here read as "titled a feast of the BVM", not verified oration-by-
+   oration, and said so explicitly rather than overclaimed. TWO entries
+   found and deliberately EXCLUDED after individual consideration, not
+   merely omitted: "dedication-of-the-basilica-of-st-mary-major" (5
+   August, Class3, Feast) -- a DEDICATION feast (of the building, "In
+   Dedicatione S. Mariae ad Nives"), whose own oration could not be found
+   in either scan under this exact heading to confirm it invokes her
+   intercession the same way an ordinary Marian commemoration does, so left
+   out per this project's "a wrong citation is worse than a missing one"
+   rule; and "purification-of-the-blessed-virgin-mary" (2 February,
+   Class2), already deliberately tagged [subject = Lord] by an earlier
+   task's own ruling (register §6.0, following the oracle's Sunday-
+   displacing treatment) for an unrelated reason -- its own RG 112(a)
+   Lord-vs-Lord exclusion already covers it against another Lord-subject
+   winner, and at Class2 it can never lose to a Class4 candidate in the
+   first place (this or any other), so including or excluding it here has
+   no live consequence either way; excluded for consistency with the
+   earlier ruling rather than silently overriding it.
+
+   What this rule COVERS: any admitted candidate on {!marian_slugs}' own
+   closed list, or tagged [subject = Bvm], losing to a WINNER that is
+   itself either {!marian_slugs}-listed or [subject = Bvm] -- currently
+   live only for the BVM Saturday Office (subject-tagged, its own slug
+   deliberately NOT on this list, {!bvm_saturday_names}'s own "Slug"
+   citation in temporal_ef.ml) as winner against Mt Carmel as loser, plus
+   (checked, not live: 24 September, "our-lady-of-ransom", the ONLY other
+   {!Celebration.status.Commemoration_only} Marian entry in the data) --
+   PROVEN structurally unreachable: 24 September falling on a Saturday
+   forces 1 September to be a Thursday (23 days = 3 weeks 2 days earlier),
+   which by {!third_sunday_of_september}'s own construction makes 24
+   September the September Ember Saturday EVERY time, RG 91 entry 18
+   (Class2), always outranking entry 27 -- not a sampled coincidence, an
+   exact day-of-week identity, checked against several sample years above
+   before being generalised. What this rule does NOT cover, stated
+   plainly: a hypothetical non-Marian-titled saint whose own oration
+   happens to invoke the BVM's intercession in passing (RG 112(d)'s own
+   text does not restrict itself to Marian-TITLED commemorations) -- no
+   oration text is stored anywhere in this codebase to detect that, and
+   none is guessed at here. Also does not cover RG 112(d)'s OTHER half (two
+   commemorations of the SAME non-BVM saint) -- unbuilt, unaffected by this
+   fix, register §4/§6 already tracks it as open.
+
+   Checked for what this fix does NOT change: {!Precedence.resolve} holds
+   [Commemoration_only] candidates out of the WINNER contest entirely, so
+   Mt Carmel could never have been [observed] either before or after this
+   fix -- only its own admission AS a commemoration changes. Verified live
+   against real 2033/2005 data (both years 16 July is a Saturday): before
+   this fix, [comms] included "our-lady-of-mt-carmel:ordinary"; after,
+   [comms] is empty and the omitted list carries it with reason "omitted:
+   yielded to a higher day", the same generic reason every other [Omit]
+   disposition in this function produces. *)
+let marian_slugs =
+  [ "annunciation-of-the-blessed-virgin-mary";
+    "assumption-of-the-blessed-virgin-mary";
+    "immaculate-conception-of-the-blessed-virgin-mary";
+    "immaculate-heart-of-mary";
+    "maternity-of-the-blessed-virgin-mary";
+    "most-holy-name-of-mary";
+    "nativity-of-the-blessed-virgin-mary";
+    "our-lady-of-lourdes";
+    "our-lady-of-mt-carmel";
+    "our-lady-of-ransom";
+    "our-lady-of-the-rosary";
+    "presentation-of-the-blessed-virgin-mary";
+    "queenship-of-the-blessed-virgin-mary";
+    "seven-sorrows-of-the-blessed-virgin-mary";
+    "visitation-of-the-blessed-virgin-mary" ]
+
+let is_bvm_office (c : Vocab_ef.rank Precedence.candidate) =
+  c.Precedence.cel.Celebration.subject = Subject.Bvm
+  || List.mem (Slug.to_string c.Precedence.cel.Celebration.slug) marian_slugs
+
 (* Task 8: what happens to the day's LOSING candidate (docs/research/
    rules-register.md §4, "Occurrence" RG 92-95 and "Vigils" RG 33, plus RG
    94; also §6.0/Caput III "De Dominicis" RG 16(a), below). [band] above
@@ -602,7 +751,16 @@ let disposition ~(winner : Vocab_ef.rank Precedence.candidate)
   let open Vocab_ef in
   let cel = loser.Precedence.cel in
   let is_temporal = loser.Precedence.origin = Precedence.Temporal in
-  if cel.Celebration.status = Celebration.Commemoration_only then
+  if is_bvm_office winner && is_bvm_office loser then
+    (* RG 112(d) -- see [marian_slugs]'s own citation above for the full
+       argument. Checked FIRST, even ahead of the [Commemoration_only]
+       branch immediately below, because the one live witness
+       (Mt Carmel, 16 July) IS [Commemoration_only] -- that branch's own
+       "always Commemorate, nothing overrides it" previously had no
+       exception for two commemorations invoking the identical BVM, the
+       exact gap this fix closes. *)
+    Precedence.Omit
+  else if cel.Celebration.status = Celebration.Commemoration_only then
     (* Always -- checked before RG 33's omission and RG 95's transfer so
        neither can override it: a Commemoration_only entry can never win
        (Precedence.resolve holds it out of the band contest entirely, see
diff --git a/test/test_differential.ml b/test/test_differential.ml
index 3f416fd..5093484 100644
--- a/test/test_differential.ml
+++ b/test/test_differential.ml
@@ -138,6 +138,7 @@ module Slug = Colitur_kernel.Slug
 module Date = Colitur_kernel.Date
 module Cel = Colitur_kernel.Celebration
 module Colour = Colitur_kernel.Colour
+module Subject = Colitur_kernel.Subject
 module V = Rite_ef.Vocab_ef
 
 (* Same relative paths test_rite_ef.ml/test_sanctoral_ef.ml use: dune test
@@ -216,6 +217,14 @@ type row = {
   slug : string;
   rank : string;
   colour : string;
+  (* Fix round 1 (coordinator finding F3): ONLY ever populated for colitur's
+     own row (see [colitur_rows_2005_2050] below) -- lectio's own fixture
+     format has no subject column at all, so [row_of_line] (the lectio-side
+     parser) sets this to "-", a placeholder never read on that side. Every
+     Layer C predicate that reads [subject] must therefore read it off the
+     COLITUR row [c], never [l] -- the same asymmetry [row]'s other fields
+     do not have, called out here rather than left implicit. *)
+  subject : string;
 }
 
 let read_lines path =
@@ -232,7 +241,7 @@ let read_lines path =
 let row_of_line line =
   match String.split_on_char ' ' line with
   | date :: weekday :: season :: week :: slug :: rank :: colour :: _others ->
-      { date; weekday; season; week; slug; rank; colour }
+      { date; weekday; season; week; slug; rank; colour; subject = "-" }
   | _ -> Alcotest.failf "malformed fixture line (fewer than 7 fields): %S" line
 
 let lectio_rows () = List.map row_of_line (read_lines fixture_path)
@@ -271,7 +280,8 @@ let colitur_rows_2005_2050 () =
               week;
               slug = Slug.to_string cel.Cel.slug;
               rank = V.rank_to_string cel.Cel.rank;
-              colour = Colour.to_string cel.Cel.colour
+              colour = Colour.to_string cel.Cel.colour;
+              subject = Subject.to_string cel.Cel.subject
             }
             :: !rows
       | None -> Alcotest.failf "internal error: no resolved day for %s" (Date.to_iso8601 !d));
@@ -535,8 +545,24 @@ let holy_name_fallback_dates_2005_2050 =
    IV-class Saturday", reduces exactly to this triple), not a short,
    individually-interesting list of dates the way C14/C15/C16 above are. *)
 let is_bvm_saturday_row (c : row) diffs =
+  (* Fix round 1 (coordinator finding F3): [c.subject] added -- the ONLY
+     field in this predicate that pins WHICH celebration is observed, not
+     merely its shape. Every other Layer C entry pins a colitur slug (C1's
+     own [jan_6_13_slug], C6's [nativity_octave_day_slugs], C8/C14/C15/C16's
+     own literal slug checks); this entry could not, because the office
+     deliberately REUSES the ordinary ferial slug (Rite_ef.Temporal_ef's own
+     [bvm_saturday_names] citation, "Slug" paragraph) -- there is no fixed
+     slug string to pin. [subject = "bvm"] is the field that DOES uniquely
+     identify the office (set nowhere else the differential's own
+     [colitur_rows_2005_2050] can produce a Saturday/Class4/white
+     combination for), closing the gap a shape-only predicate left open:
+     without this conjunct, a FUTURE bug that made some OTHER white,
+     Class4, Saturday candidate exist (Christmastide/Paschaltide, where
+     [season_colour] is already white, so a real bug there could slip
+     through unnoticed by colour alone) would be silently absorbed here
+     too. *)
   diffs = [ Colour_f ] && String.equal c.weekday "saturday" && String.equal c.rank "class-4"
-  && String.equal c.colour "white"
+  && String.equal c.colour "white" && String.equal c.subject "bvm"
 
 (* [layer_c_reason l c diffs] returns the [data/ef/expected-divergences.sexp]
    [id] this row-pair's remaining (post Layer A/B) diff set belongs to, or
diff --git a/test/test_golden.ml b/test/test_golden.ml
index bb4bcfc..fdfc01d 100644
--- a/test/test_golden.ml
+++ b/test/test_golden.ml
@@ -940,6 +940,24 @@ let test_bvm_saturday_loses_to_most_holy_name_of_mary_2026 () =
      colour=white subject=bvm name_la=- comms=[] in=- out=[]"
     (describe d)
 
+(* RG112(d), fix round 1 (coordinator finding F1): PE.marian_slugs's own
+   citation (precedence_ef.ml) has the full RG112(d) argument, including
+   the 16 July "Missa dici potest aut ... aut ..." rubric found on both
+   photographic scans. 16 July 2033 is a real otherwise-unoccupied
+   Saturday; before this fix, [comms] here read
+   "[our-lady-of-mt-carmel:ordinary]" (confirmed against real pre-fix
+   `colitur day 2033` output). *)
+let test_bvm_saturday_excludes_mt_carmel_2033 () =
+  let d = fetch 2033 7 16 in
+  Alcotest.(check bool)
+    "2033-07-16: Mt Carmel is in [omitted] (RG112(d)), not silently dropped from the record" true
+    (omitted_has d "our-lady-of-mt-carmel");
+  Alcotest.(check string)
+    "2033-07-16: the BVM Saturday office observed, white; NO commemoration of Mt Carmel (RG112(d))"
+    "2033-07-16 saturday season=time-after-pentecost week=5 slug=ef-time-after-pentecost-5-saturday \
+     rank=class-4 colour=white subject=bvm name_la=Officium sanctae Mariae in sabbato comms=[] in=- out=[]"
+    (describe d)
+
 let suite =
   ( "golden pins (known-tricky years)",
     [ Alcotest.test_case "Easter extreme: 1598 earliest (22 Mar, Gauss-verified)" `Quick
@@ -1004,5 +1022,9 @@ let suite =
       Alcotest.test_case
         "RG91 e27/RG78/RG26: the BVM Saturday office loses to Most Holy Name of Mary (subject Bvm too) \
          and is omitted, no confusion (2026-09-12)"
-        `Quick test_bvm_saturday_loses_to_most_holy_name_of_mary_2026
+        `Quick test_bvm_saturday_loses_to_most_holy_name_of_mary_2026;
+      Alcotest.test_case
+        "RG112(d), fix round 1 (F1): the BVM Saturday office excludes Mt Carmel's own commemoration \
+         (2033-07-16)"
+        `Quick test_bvm_saturday_excludes_mt_carmel_2033
     ] )
diff --git a/test/test_precedence_ef.ml b/test/test_precedence_ef.ml
index 475ae6e..6f8b5b5 100644
--- a/test/test_precedence_ef.ml
+++ b/test/test_precedence_ef.ml
@@ -805,6 +805,52 @@ let disposition_cases =
         "conversion-of-st-paul",
       cand ~origin:P.Sanctoral ~rank:V.Class2 ~subject:Sub.Lord ~layer:PE.universal_layer
         "commemoration-of-the-baptism-of-the-lord",
+      "Commemorate(Ordinary)" );
+    (* RG112(d), fix round 1 (coordinator finding F1) -- PE.marian_slugs's
+       own citation has the full argument (both scans, including the 16
+       July "Missa dici potest aut ... aut ..." rubric). [of_temporal] on a
+       REAL 16-July-Saturday (2033) sources the winner from
+       Temporal_ef.temporal's own real output, the same review-finding-3
+       discipline this file's Ember/vigil rows already use -- not a
+       hand-typed slug/subject pair that could silently drift from what the
+       office actually builds. The loser is the real data/ef/sanctoral.sexp
+       entry, [Commemoration_only], the exact shape that made this branch
+       need to be checked BEFORE the Commemoration_only catch-all. *)
+    ( "RG112(d): the BVM Saturday Office excludes a commemoration invoking \
+       the SAME BVM's intercession -- Omit, not Commemorate (Mt Carmel, \
+       16 July 2033, a real otherwise-unoccupied Saturday)",
+      of_temporal (mk 2033 7 16),
+      cand ~origin:P.Sanctoral ~rank:V.Class3 ~status:Cel.Commemoration_only ~subject:Sub.Saint
+        ~layer:PE.universal_layer "our-lady-of-mt-carmel",
+      "Omit" );
+    (* Conjunct 1/2, LOSER identity: an ORDINARY Commemoration_only saint
+       NOT on PE.marian_slugs must still reach the ordinary commemorate
+       flow even when the winner is the BVM Saturday Office -- proves
+       RG112(d) excludes a specific same-BVM collision, not every
+       commemoration the office happens to outrank (the real live shape,
+       1 August, Holy Machabees, PE.marian_slugs's own comment). *)
+    ( "RG112(d) does NOT fire when the LOSER is not a Marian commemoration \
+       -- an ordinary saint stays Commemorate(Ordinary) even under the BVM \
+       Saturday Office (1 August 2026, Holy Machabees)",
+      of_temporal (mk 2026 8 1),
+      cand ~origin:P.Sanctoral ~rank:V.Class3 ~status:Cel.Commemoration_only ~subject:Sub.Saint
+        ~layer:PE.universal_layer "holy-machabees",
+      "Commemorate(Ordinary)" );
+    (* Conjunct 2/2, WINNER identity: an ordinary (non-Bvm, non-Marian-
+       titled) winner must not exclude Mt Carmel either -- RG112(d) is
+       keyed on the WINNER also being "de B. Maria Virg.", not merely on
+       the loser's own identity. SYNTHETIC: no live date has an ordinary
+       Class3 saint winning outright while Mt Carmel also loses that same
+       day (Mt Carmel's own fixed date, 16 July, has no other real
+       sanctoral entry to construct this from) -- proves the conjunct is
+       real, not merely untested, the same "SYNTHETIC" discipline the
+       RG112(a) rows above already use for their own second conjunct. *)
+    ( "SYNTHETIC: RG112(d) does NOT fire when only the LOSER is Marian -- \
+       an ordinary winner leaves Mt Carmel Commemorate(Ordinary)",
+      cand ~origin:P.Sanctoral ~rank:V.Class3 ~subject:Sub.Saint ~layer:PE.universal_layer
+        "ef-synthetic-ordinary-winner",
+      cand ~origin:P.Sanctoral ~rank:V.Class3 ~status:Cel.Commemoration_only ~subject:Sub.Saint
+        ~layer:PE.universal_layer "our-lady-of-mt-carmel",
       "Commemorate(Ordinary)" )
   ]
 
@@ -1507,6 +1553,32 @@ let test_bvm_saturday_loses_to_a_real_feast_end_to_end () =
     [ ("ef-time-after-pentecost-15-saturday", "omitted: yielded to a higher day") ]
     (List.map (fun (c, reason) -> (S.to_string c.P.cel.Cel.slug, reason)) resolution.P.omitted)
 
+(* RG112(d), fix round 1 (coordinator finding F1) -- PE.marian_slugs's own
+   citation has the full argument. End-to-end through the REAL pipeline
+   (PE.band, PE.disposition AND PE.admit together): 16 July 2033, a real
+   otherwise-unoccupied Saturday, with the real data/ef/sanctoral.sexp
+   [our-lady-of-mt-carmel] entry as the day's only competing candidate.
+   Before this fix, [resolution.P.commemorations] held
+   [our-lady-of-mt-carmel:Ordinary] (confirmed live against real 2005/2033
+   `colitur day` output before this fix existed); after, it is empty and
+   Mt Carmel is reported omitted, not silently dropped. *)
+let test_bvm_saturday_excludes_mt_carmel_end_to_end () =
+  let date = mk 2033 7 16 in
+  let day_ctx = ctx date in
+  let bvm_saturday = of_temporal date in
+  let mt_carmel =
+    cand ~origin:P.Sanctoral ~rank:V.Class3 ~status:Cel.Commemoration_only ~subject:Sub.Saint
+      ~layer:PE.universal_layer "our-lady-of-mt-carmel"
+  in
+  let resolution = P.resolve real_rules day_ctx ~temporal:bvm_saturday ~sanctoral:[ mt_carmel ] in
+  Alcotest.(check string) "the BVM Saturday office is observed"
+    "ef-time-after-pentecost-5-saturday" (S.to_string resolution.P.observed.P.cel.Cel.slug);
+  Alcotest.(check (list string)) "Mt Carmel is NOT commemorated (RG112(d))" []
+    (List.map (fun (c, _) -> S.to_string c.P.cel.Cel.slug) resolution.P.commemorations);
+  Alcotest.(check (list (pair string string))) "Mt Carmel is reported omitted, not silently dropped"
+    [ ("our-lady-of-mt-carmel", "omitted: yielded to a higher day") ]
+    (List.map (fun (c, reason) -> (S.to_string c.P.cel.Cel.slug, reason)) resolution.P.omitted)
+
 (* RG 16(a) (register §6.0) reached end-to-end through the REAL pipeline
    ([PE.band], [PE.disposition] AND [PE.admit] together, not any one of
    them in isolation as this file's own hand-tagged rows test them
@@ -1989,6 +2061,10 @@ let suite =
            (2026-09-12)"
           `Quick test_bvm_saturday_loses_to_a_real_feast_end_to_end;
         Alcotest.test_case
+          "RG112(d), fix round 1 (F1): the BVM Saturday office excludes Mt Carmel, end-to-end \
+           (2033-07-16)"
+          `Quick test_bvm_saturday_excludes_mt_carmel_end_to_end;
+        Alcotest.test_case
           "RG16(a): a Feast of the Lord suppresses the Sunday's own commemoration entirely, \
            end-to-end (2028-08-06)"
           `Quick test_rg16a_lord_feast_suppresses_sunday_end_to_end;
diff --git a/test/test_temporal_ef.ml b/test/test_temporal_ef.ml
index ff3cc0b..4e97d33 100644
--- a/test/test_temporal_ef.ml
+++ b/test/test_temporal_ef.ml
@@ -507,25 +507,40 @@ let test_colours () =
    (6 Jan) is a Tuesday, so the actual first-Sunday-after-Epiphany origin is
    11 Jan -- exercising all four sub-stretches of the fix. *)
 let test_christmastide_feria_slugs () =
-  (* 26 Dec 2026 and 10 Jan 2026 are both SATURDAYS -- ef-bvm-saturday task
-     (RG 78, "De sancta Maria in sabbato"): every Christmastide feria
-     [christmastide_feria_slug] builds is unconditionally Class4
-     ([ferial_rank]'s own catch-all outside Advent/Lent/Passiontide), so on a
-     Saturday {!Rite_ef.Temporal_ef.bvm_saturday_names}'s own override fires
-     here too -- but per that citation's own "Slug" paragraph, the SLUG is
-     deliberately left UNCHANGED (reused from this same "-0-"/"-2-" naming,
-     not replaced): only colour/subject/name change. So the slug assertions
-     below are unaffected by the fix; what is NEW is the colour/subject
-     override on these two specific dates, asserted alongside them. *)
+  (* 26 Dec 2026 is a SATURDAY -- but, fix round 1 (coordinator finding F2),
+     it can NEVER actually hold the BVM Saturday office in the real
+     resolved pipeline: 26/27/28 December are St Stephen/St John/the Holy
+     Innocents (RG 67-70), fixed, unconditional, real data/ef/sanctoral.sexp
+     entries every single year (rank >= Class2), which always outrank a
+     Class4 temporal candidate -- so RG 78's own protasis ("in quibus
+     OCCURRIT Officium de feria IV classis") is never satisfied by ANY day
+     in the 26-28 December stretch, in any year. [T.temporal] itself has no
+     visibility into the sanctoral layer at all (by design -- the temporal
+     candidate is computed independent of what saint's day it is; Precedence
+     decides the winner afterward), so it still unconditionally tags this
+     date [subject = Bvm] -- a genuine, correct fact about [T.temporal]'s
+     own pure computation, but NOT a claim that this shape is ever the real
+     OBSERVED office, and pinning it with RG-78 citation language as if it
+     were would overclaim. Only the SLUG is asserted here (unaffected
+     either way, unconditionally correct); the override demonstration
+     itself uses 3 January and 10 January below instead -- both, checked
+     directly against data/ef/sanctoral.sexp, have no competing entry in
+     any year, so the BVM Saturday office genuinely IS reachable there. *)
   Alcotest.(check string) "26 Dec (Sat)" "ef-christmas-0-saturday" (slug_of (d 2026 12 26));
-  Alcotest.(check string) "26 Dec (Sat): the BVM Saturday office overrides the season colour (white, RG 120(b))"
-    "white" (colour_of (d 2026 12 26));
-  Alcotest.(check bool) "26 Dec (Sat): subject Bvm" true (subject_of (d 2026 12 26) = Sub.Bvm);
   Alcotest.(check string) "28 Dec (Mon)" "ef-christmas-0-monday" (slug_of (d 2026 12 28));
+  (* 3 Jan 2026 is a SATURDAY, genuinely reachable (data/ef/sanctoral.sexp
+     has no 3 January entry) -- the real demonstration for the "-1-"
+     stretch that 26 December's own Saturday could not honestly provide. *)
+  Alcotest.(check string) "3 Jan (Sat): slug stays the ordinary \"-1-\" naming" "ef-christmas-1-saturday"
+    (slug_of (d 2026 1 3));
+  Alcotest.(check string) "3 Jan (Sat): the BVM Saturday office overrides the season colour (white, RG 120(b))"
+    "white" (colour_of (d 2026 1 3));
+  Alcotest.(check bool) "3 Jan (Sat): subject Bvm" true (subject_of (d 2026 1 3) = Sub.Bvm);
   Alcotest.(check string) "2 Jan (Fri)" "ef-christmas-1-friday" (slug_of (d 2026 1 2));
   Alcotest.(check string) "5 Jan (Mon)" "ef-christmas-1-monday" (slug_of (d 2026 1 5));
   (* 7-10 Jan precede the actual origin Sunday (11 Jan): colitur-only, not
-     lectio's "week 1" key, precisely to avoid colliding with 12-13 Jan below. *)
+     lectio's "week 1" key, precisely to avoid colliding with 12-13 Jan below.
+     10 January is ALSO genuinely reachable (no sanctoral entry that date). *)
   Alcotest.(check string) "7 Jan (Wed, before the origin)" "ef-christmas-2-wednesday" (slug_of (d 2026 1 7));
   Alcotest.(check string) "10 Jan (Sat, before the origin)" "ef-christmas-2-saturday" (slug_of (d 2026 1 10));
   Alcotest.(check bool) "10 Jan (Sat): subject Bvm" true (subject_of (d 2026 1 10) = Sub.Bvm);