Symmetric fits leave the mean eclipse offset unresolved in one July 2025 JWST visit of WASP-103 b

Two secondary eclipses bracket a primary transit in JWST program 5022, observation 1 of WASP-103 b. We fit symmetric event models to public NIRSpec PRISM band data from a verified SCI source. We measure D, the mean eclipse center minus the transit center. A common additive offset in the three reported centers cancels from D. The main white sum gives D = +6.0 seconds. Residual-block refits give standard deviations of 27.5 seconds for 5-minute blocks and 21.8 seconds for 20-minute blocks. All 18 tested white models give D from +3.2 to +16.0 seconds. These fits do not resolve a nonzero mean offset. An exploratory band comparison gives D(2.8–3.8 micrometres) minus D(2.0–2.8 micrometres) = -27.1 seconds. Its paired block standard deviations are 10.5 and 8.6 seconds. All 13 tested spectral models retain this difference, from -35.4 to -25.4 seconds. The pair was selected after inspection. All six infrared pairs and all model controls are retained. The block intervals are conditional and have no established coverage for this observation. Primary limb parameters often reach bounds. Numerical checks and synthetic noise tests show the limits of formal covariance errors. The results do not establish an atmospheric cause, orbital eccentricity, or period change. The package preserves code, copied-input hashes, data provenance, all fit and block results, five figures, and reproduction instructions.

WASP-103 b · JWST · secondary eclipse · transit timing · NIRSpec PRISM · correlated noise · reproducible analysis