Divides profit by COST, not revenue — a markup ratio (~138%) wearing margin's label; verified margin is 57.94% of revenue
2
Fix
12.0% growth is verified TRUE; "the West's 112.7% growth" is FALSE — 112.7% is the West's target attainment, not growth. Half-true prose is the hardest fabrication class
3
Accept
Correct composition; verified 3.033
4
Reject (rebuild)
Eleven equal visuals with no hierarchy fails the L10 pass at the question test — before alignment is even discussed
5
Fix
"Top seller" is ambiguous: top REVENUE product, yes ($867.6K); by units it ranks nowhere near Cedarline's 1,399. The three-crowns discipline: name the measure
6
Reject
Averages returned quantities per return line (~1.26) — not a rate of anything; verified Return Rate is 3.21% (returned units over sold units)
Reflection answers
Artifacts 3 and… only 3 is cleanly correct as labeled — 2 and 5 carry true cores
needing fixes, which is the trick: the gauntlet has ONE clean accept, and recognizing
that 5's claim needed a measure named is part of clearing honestly. Clearing fast and
plainly is the same discipline as the chart review and the checker philosophy: precision
is what makes your rejections credible.
Artifact 2 is hardest because a verified true number sits beside a mislabeled one —
the reader's guard drops after the first check passes. The rule: verify each
claim in generated prose independently; truth is not contagious between adjacent
sentences, and 112.7% was real — it was just the answer to a different question.