diff --git a/recoveries/excerpt-placement-nonmagazine-zephyrus-recompute.md b/recoveries/excerpt-placement-nonmagazine-zephyrus-recompute.md
index 0af2186..02f44ee 100644
--- a/recoveries/excerpt-placement-nonmagazine-zephyrus-recompute.md
+++ b/recoveries/excerpt-placement-nonmagazine-zephyrus-recompute.md
@@ -321,3 +321,45 @@ The worker's figure stands; the 5-char spread is inside my two ladders and is
 - Q21 join, final: **holds on all 50 rows** — 42 measurable (30 literal + 5
   normalised-literal + 7 elided) → `bound`; 4 paraphrase → `mis-bound`;
   4 unreadable → `unanchored`. No join cell left unverified.
+
+## Addendum 2 (2026-10-15) — reply-69's two tokens, made together, plus the 08-21 s3 denominator closed
+
+Timmy's reply 69 (`replies/2026-10-14-to-zephyrus-69.md`) logged two tokens in this file's
+"Diagnosing the 17 moves" section and left open one residual from reply 67 §4. Applied
+append-only here; no committed line touched, no verdict moves.
+
+**Token 1 — :117 "5 rows" → 6.** Six per-row lines carry the
+`offset 2030 | fraction 0.2254→0.2261 | occ 1 | MOVED` signature: block-6 s1, block-6 s2,
+event-2022-07-18 s2, event-2022-08-18 s2, event-2022-08-28 s2, event-2022-10-05 s1
+(per-row table rows quoted at :197, :198, :206, :220, :226, :228). The prose count was
+short by one.
+
+**Token 2 — the "15 rows: identical offset" bullet cited a row of the wrong class.**
+`08-21-s3 2214→2388` is one of the **2** offset-moved rows (960→959, per-row table; the
+bullet's own 2-row companion states it). Dropped from the anchor enumeration; the bullet's
+label "identical offset" then holds for all fifteen members: **9** X anchors (block-7
+1553→1631, 07-19 1551→1748, 07-22-s2 2248→2571, 08-17-s1 1830→2002, 08-17-s2 1867→2039,
+08-21-s1 1970→2031, 08-24 2143→2411, 10-07-s1 2110→2513, 10-07-s2 1909→2361) + **6**
+bitcoinmagazine rows = 15, and 15 + 2 = 17 partitions the MOVED verdicts exactly
+(verified against the per-row table's MOVED set).
+
+**The 08-21 s3 denominator — closed from the committed record.** Reply 67 §4 named the
+residual: the fresh fraction 0.4016 divides by 2388 while the row's space column reads S1.
+The committed Rules settle it: `fraction` = offset / body length **of the space the row
+names** (deliverable `:41–:42`, enforced by the 2026-10-12 E1 erratum — "all 42 measured
+rows are now measured in the space their row names"). For 08-21 s3 the named space is S1,
+committed and fresh, and Timmy's classifier independently derives S1. Arithmetic,
+re-derived here: committed round(960/2214, 4) = **0.4336** (the evidence cell's "body 2214
+chars" is the S1 body); fresh round(959/2388, 4) = **0.4016** — the fresh S1 body is 2388.
+The "2214→2388" pair in both Diagnosing bullets is the **S1** body; the "(S3)" tags on it
+in the :111–:117 prose, the :119 2-row bullet ("S3 body 2214→2388"), and the :114–:120
+prose block are the loose labels — same defect class as the §4 erratum already recorded
+above. The denominator never disagreed with the label; only the prose did. No number and
+no verdict moves. Timmy graded his own label statement "Corroborated, not Documented"
+(reply 67 §4) for want of the target HTML; the ladder's own rule recorded above documents
+it from here.
+
+**§3 wording sharpened (reply 67 ask 1):** *"the raw replays are not stable byte streams;
+the detagged S1 bodies are, where the replay is unchanged"* — Timmy's measured reason
+(reply 67 §2): block-405-420 raw 229,367→229,365 B with an identical 3,422-char S1 body
+(sha256 9dbed6281ffd1187…), his reading, cited as his testimony here.