One sample accounts for the 4.15 percent flux contrast in the JWST 1.2–1.8 micrometre sum with equal weights for WASP-12 b on 10 March 2026

We test two flux states in the public JWST NIRSpec PRISM data for WASP-12 b on 10 March 2026. We use the delivered calibrated product with 26,100 integrations. The 1.2–1.8 micrometre band sum and a 2.0–2.8 micrometre reference sum use equal spectral weights. Sample 81, at 1.3217885 micrometres, has a reported extraction count of four or seven in 26,072 records. Its four-pixel state has about 92.2 percent less flux. All 4,815 four-pixel records have a delivered data quality flag of zero. Outside transit, the band/reference ratio has a 41,507 ppm state contrast. Removal of this one sample at every integration reduces the contrast to −80 ppm. Separate early and late checks, adjacent transitions, neighboring masks, and frequency-width weights support the localization. The four-pixel state is frequent outside transit and nearly absent in the core. With fixed transit geometry, the main maximum-template-depth estimate changes from 2,488 to 14,962 ppm after the mask. Eighteen tested fits per band give depth ranges of 1,329–3,888 and 14,374–15,802 ppm. These ranges measure model sensitivity. The masked fit retains excess residual noise. The result locates a calibrated-product effect and tests a fixed mask. The detector cause and a calibrated physical depth remain unresolved. We preserve reduced native data, code, archive hashes, figures, and an exact raw-data reproduction.

WASP-12 b · JWST · NIRSpec PRISM · extraction pixels · data quality · transit depth · wavelength mask · public data · reproducible research