Seasonal activity subtraction removes 55 percent of a wave's energy near 34.9 days in a fixed GJ 674 test

We test signal loss from linear activity subtraction with 224 GJ 674 HARPS exposures in five dense seasons. Exact phase bounds use the sampled H-alpha, differential line width and chromatic index. With separate activity coefficients per season and a 2 m/s extra error scale, one wave near 34.9 days loses 54.9 percent of its weighted energy after baseline removal. Common three-index and H-alpha fits lose up to 39.0 and 30.3 percent on the tested grid. Three weight settings, a finer period grid, direct injections, and an independent least-squares audit support the conditional result. Joint fitting recovers injected coefficients but does not restore information lost to overlap with activity. This is a search-method test. It does not detect a new planet or identify a physical cause. Fixed measured indices, shared spectra and unknown covariance limit interpretation. Code, aggregate results, source hashes, a figure and repeat steps are included. External source rows remain outside the bundle.

astronomy · GJ 674 · HARPS · activity correction · signal attenuation · planet search · linear projection