95% bounds exclude a common four-band correction for 355 Gaia and SMASS asteroid pairs under the stated Gaussian model

We test whether one positive multiplier at each of four Gaia wavelengths can bring 355 asteroid contrasts into agreement with SMASS II. The fixed sample has Bus classes B, C, Cb, Cg, Ch, and Cgh. We assume centred Gaussian marginal errors with fixed listed standard deviations and nonnegative latent reflectance. We permit any error dependence across wavelengths, surveys, and objects. A union bound gives simultaneous 95% native intervals. The correction has c550 = 1 and unrestricted nonnegative coefficients at 682, 726, and 858 nm. Ratio overlap gives 710 linear inequalities. Three pairs, 52 Europa, 144 Vibilia, and 423 Diotima, give an exact certificate that no common correction satisfies these bounds. Doubling all listed errors permits a positive correction. Thus, the result depends on the measurement model. A separate correction fitted to 177 objects reduces mean absolute contrast disagreement by 22.9% in 178 other objects. The mean error falls from 1.989 to 1.533 percentage points. An additive training offset gives 1.522 percentage points, with no clear advantage for the four-band method. The study extends an earlier additive-offset test to the whole four-band correction space. It does not test published external correction coefficients or identify the cause of the survey difference. We preserve independent exact verification, source requests and hashes, code, diagnostic results, figures, and reproduction instructions.

asteroids · Gaia DR3 · SMASS II · reflectance spectra · survey calibration · simultaneous confidence bounds · unknown error dependence · linear inequalities · held-out comparison · reproducible research