Local eclipse fits reduce artificial flare recovery in 25 TESS sectors of CM Draconis

We tested how local eclipse fits affect flare detection in 25 public TESS sectors of CM Draconis. We added 7,200 template flares to two-minute light curves and refit the trend and local eclipses. At a true template scale of 1% of system flux, recovery above a measured 0.5% cut was 27.1% during eclipse and 77.1% at other phases. In 200 paired controls, holding the original model fixed gave 93 recoveries from 100 eclipse injections. Full-sector refits gave 47. The corresponding counts outside eclipse were 87 and 80. Local and full-sector fits gave the same recovery decisions in these controls. For 50 trials across all sectors, refitting the sector shape also gave the same decisions. A model decomposition shows that eclipse fits absorb most of the lost peak signal in those tests. The observed catalog has 255 candidates, including 92 in ten sectors from 2024. Nineteen candidates peak during eclipse. The observed eclipse-to-other rate ratio is 0.834, with an exact sector-conditioned 95% interval of 0.493 to 1.332. This interval describes detected candidates under an independent count model. The recovery grid does not determine the real flare population, so it cannot give a unique intrinsic rate correction. The package preserves input arrays, source hashes, code, candidate and injection tables, figures, and reproduction instructions.

CM Draconis · TESS · stellar flares · eclipsing binaries · injection recovery · selection effects · public data