DC2018: the C29 caveat was wrong about where the convention residual lands

[jdeast]

ab194/av_true — the first run with the colour-anchored av prior (9.01 ± 0.3, ptde, Rhat 1.00, ESS 23–25k, 78/78 chains in the good region) — contradicts the caveat dc18_truth_table.py prints, so this rewrites it rather than leaving it in place.

What it claimed, and what was measured

The caveat said theta_star is biased low by 10**(-0.2*grey) = 0.766x on event 194, read straight off the challenge's own colour–surface-brightness relation. Measured: 0.983x.

The arithmetic is right about the CSB relation and wrong about the target. theta_star is protected, because the light curve pins rho * theta_E independently of the SED. Cross-checked: rho = theta_star / theta_E reproduces both the truth and the fitted rho from the scored table.

Where it actually lands

quantity	truth	fit	fit/truth	pull
theta_star (= R/D)	3.829e-4 mas	3.763e-4	0.983	—
R_source	0.822 R☉	0.331	0.403	−5.25
D_source	9980 pc	4090	0.410	−5.35
theta_E	0.3486 mas	0.27	0.774	−0.66

R_source and D_source are each ~2.45x low but slide in lockstep, so their ratio — which is theta_star — survives while each is individually −5 sigma. The flux the grey term over-predicts is absorbed by a source made too cool (teffsed 3040 K here), and the distance then follows the resulting luminosity inward.

The prior fix is a real win the old text buried

theta_star vs truth   0.62x  (old arms, av ~ 4.25 written from A_W149)
                ->    0.983x (colour-anchored, av = 9.01)
R_source pull        -13.10  ->  -5.25

What the new text is careful about

• It says explicitly not to read the surviving −5 sigma pulls as convention-explained. That is not established.
• It drops the per-event theta_star-bias column, which was the wrong quantity to tabulate.
• A truth note, because I had it backwards in an earlier report: event 194's source is a dwarf (0.822 R☉ at 9980 pc), not a clump giant, so a fitted 0.331 R☉ is 2.45x low rather than the factor of 30 a giant comparison implies.

The matching correction is in the notes repo (paper_extinction_convention.txt, section 4b), and issue #303 carries a comment so the paper agent does not write the retracted version.

Docs/script only — no library code, no test fixtures.