Excluding two BAT days reverses the 2024–2026 cubic and sine clock ranking for LMC X-4 with fixed four-harmonic profiles and inverse-variance fits

We test the central cubic and sine phase clocks in Chou (2026, arXiv:2602.13491v1) on later public LMC X-4 data. We fit four-harmonic profiles before 2024 June 1 and fix them for the later test. In the BAT calibration interval from 2017 September to 2024 May, two flag-zero days supply 90.783% of the weight from the listed errors. All 2,196 accepted days give a cubic-to-sine weighted mean squared error ratio of 1.004631 on 587 later days, through 2026 October 19. Excluding only those two calibration days changes the ratio to 0.949031. Both later prediction errors fall. Equal calibration weights and a variance floor give similar errors. This exclusion is a check made after the first fits. It does not show that the two measurements are invalid. The original small ranking difference has a block-resampling interval that includes equal errors. The MAXI comparison also does not give a stable clock choice. Clock parameter uncertainty and profile evolution limit a physical interpretation. The contribution is a frozen later-data benchmark and a reproducible weight-influence result. The package includes exact source data, provenance, code, all controls, daily predictions, figures, and a separate numerical audit.

LMC X-4 · superorbital cycle · Swift BAT · MAXI · phase prediction · inverse-variance fitting · weight concentration · data provenance