diff --git a/docs/re/f6-unit5-pteff03a-never-drawn.md b/docs/re/f6-unit5-pteff03a-never-drawn.md
new file mode 100644
index 00000000..81b8e0a1
--- /dev/null
+++ b/docs/re/f6-unit5-pteff03a-never-drawn.md
@@ -0,0 +1,69 @@
+# F6 unit 5 — `pteff03a` is **never drawn**: the game shows ONE sweep, not two
+
+**Status: ✅ measured ⟨capture⟩ ×1.** 2026-09-02.
+
+**Question:** does `pteff03a` appear anywhere in the capture?
+**Look at:** any quad travelling right-to-left at the larger scale. Pass = found;
+fail = absent throughout.
+**Not covered:** why the disc declares it if it is absent; other screens.
+
+---
+
+## The measurement
+
+Every quad taller than 1.2 NDC in the whole capture — frames 1…2499, the title's
+build-in and settle:
+
+```
+h=3.15   n=1754   frames 746..2499   x drift +2.19   LEFT->RIGHT     <- pteff03
+h=2.00   n=  54   frames 729.. 782   x drift  0.00   static          <- not a sweep
+
+quads in pteff03a's expected height band (>3.4): NONE
+```
+
+`pteff03a` declares **800 % scale** against `pteff03`'s 600 %, so its AABB is
+≈ 1303 px ≈ **3.62 NDC**, and it travels **right-to-left**. There is no quad of
+that height anywhere, and only **one** travelling quad in the capture at all.
+
+> **The game draws one sweep during the title. The second is declared and not
+> submitted.**
+
+📌 **Not a windowing artefact.** `ptloop02`'s leaf reaches α = 128 at its `t = 150`;
+at the leaf rate measured in unit 3 that is ≈ 324 frames after its start, well
+inside the 1 754 frames captured. If it shared `pteff03`'s clock it had ample room
+to appear.
+
+## What it hands to the port
+
+Their renderer **draws** `pteff03a` — confirmed on their side at three instants.
+**An extra light streak the game never shows, appearing partway through the
+build-in, looks exactly like "the sweep starts too early."** That is a better
+candidate for the human's report than any clock offset, and it does not depend on
+unit 2's unresolved alpha question at all.
+
+⚠️ **One capture, one boot, one window.** Absence here is strong — the scan was
+over every tall quad in the whole stream, not a filtered subset — but it is still
+absence in one observation. The five-second human check the port is asking for
+(*one streak or two?*) would settle it faster than another capture, and I would
+wait for that before anyone deletes an element.
+
+## 🔴 The disc declares it, so the question moves to my side
+
+`GP_TITLE` build 4 declares `ptloop02.rat` as element 12, beside `ptloop01.rat` as
+element 11 — identical pivots, identical keyframe times. **Why the game submits one
+and not the other is not answered here.** Candidates, none tested: its leaf alpha
+starts at 0 where `pteff03`'s starts at 255, so a zero-alpha skip would suppress
+its first frames but not the whole run; or a focus/variant link means only one of
+the pair is ever active.
+
+## Refutation attempt
+
+**Target:** the port's refutation of my by-size identification — *"both sweep
+sprites are byte-identical in size, so size cannot separate them"*.
+
+**Result: they refuted it themselves first, and correctly.** The two *sprites* are
+399×180, but the two *leaves* declare different scales, so the drawn quads differ
+in height by a third. Size separates them fine. Recorded because it is the second
+time in this exchange that a challenge was right to be raised and wrong in its
+stated reason — and both times the check still found something real. **The value
+was in asking, not in the diagnosis.**