The evidence for a 1.7-day signal in March–October 2026 Her X-1 MAXI data depends on the error model

We test the known binary harmonic of Her X-1 during the late part of its 2025–2026 extended low state. The fixed interval starts at MJD 61114.9, as in prior work. A frozen public MAXI 2–20 keV table supplies 1,920 valid orbit bins from 24 March to 14 October 2026. A weighted sine and cosine fit with separate 35-day means gives a semi-amplitude of 0.00436 photons per square centimetre per second. A joint sign test over seven time blocks gives p = 0.015625. A test with independent Gaussian errors equal to the archive errors gives p = 0.12493 for the same coefficients. Tests with 17.5-day and 70-day blocks give p = 0.05322 and 0.125. The fitted amplitude is stable across local-mean, clock and weight controls, but the evidence does not establish a detection across the tested error assumptions. The approximate 35-day block ellipse gives a nominal 95% amplitude bound of 0.01035. An exploratory early-minus-late projection gives an absolute flux difference of 0.01863 and block p = 0.03308; the 70-day control gives p = 0.05469. Mean flux also falls, so this does not establish a change in fractional modulation. Direct fits, injected signals, timing checks and a limited cadence transfer test audit the calculation. The package preserves exact public inputs, hashes, code, results, figures and instructions. It does not identify a new clock or a unique emission process.

Her X-1 · MAXI · X-ray binary · orbital modulation · extended low state · cluster sign test · error model · reproducible research