Correlated noise weakens the 49.06-day circular orbit test in two public TOI-125 HARPS reductions

We test a proposed 49.06034-day transit clock at TOI-125 with two public reductions of the same 124 HARPS exposures. We remove seven exposures from three nights with a known calibration lamp fault. We then fit 62 nightly means. The model includes the three known planets, a quadratic drift, and a circular orbit at the proposed transit phase. Under white noise, the velocity coefficient K is 1.79 +/- 0.76 m/s in the CDS table. It is 1.81 +/- 0.51 m/s in the DACE G9 reduction. These are conditional standard errors. Selected correlated noise models increase the errors to about 0.7 to 1.4 m/s and give weaker support for the orbit. Removing two late observations also lowers K. Thus, these data do not securely confirm or reject a planet. The noise grid is a sensitivity test. It does not give a full noise posterior or a calibrated detection test. The transit source remains unresolved. A separate audit tests both columns labelled FWHM and BIS in the frozen CDS table J/A+A/693/A4. Both columns equal sqrt(2) times the velocity error in all 7,502 rows for 51 stars. These columns cannot supply independent line-shape measurements. Measured DACE indicators are preserved as a check. This result does not establish which indicators were used in the source paper's internal fits. The publication includes frozen public inputs, download records, source hashes, code, numerical checks, figures, and reproduction instructions.

TOI-125 · HARPS · radial velocity · transit timing · stellar activity · data audit · DACE · VizieR · correlated noise · reproducible research