Removing two BAT warning dates reduces the Cyg X-3 profile difference in a 2012–2026 three-harmonic fit

We compared the 15–50 keV orbital profiles of Cyg X-3 in low and high 30-day brightness groups. The question was whether a free amplitude could account for their difference. We used public Swift/BAT pointings from 2012 January to 2026 October. Classification used different days from each measured profile. The fit used daily constants, three harmonics, bin integration, and statistical-error weights. A large initial shape difference depended on 27 pointings with DATA_FLAG = 0. They fell on 2019 March 29 and 30, dates covered by an existing BAT data warning. These points carried 76.2% of the raw statistical weight. Their removal reduced the normalized profile distance from 0.556 to 0.045. A 90-day block null test changed from p = 0.000333 to p = 0.369. Retaining the original labels gave the same cleaned coefficients. The cleaned first-harmonic phase difference was −0.00458 cycles. Its 95% block interval was −0.01269 to +0.00340 cycles. These values are about −79 seconds, with an interval of −219 to +59 seconds. A residual simulation that repeated classification gave a shape p value of 0.406. This evidence concerns the tested full-record fit. It does not prove a constant physical profile. The later-epoch phase check remains offset. A separate audit found 8,088 integer date fields one day later than a coarse mission-time day. We used the FITS time coordinate. The package preserves frozen public inputs, provenance, code, null draws, figures, and reproduction instructions. The new result is the measured effect of known problem dates on a defined orbital profile test. It is not a planet detection or a test of a third-body model.

Cyg X-3 · Swift BAT · orbital modulation · data quality · X-ray binaries · block bootstrap · timing · reproducible analysis