Do not use the old figures to judge the 2.1 GTI strategy.
The 2026-09-02 event–GTI audit proved that the
2.filter/allevc.fits used by the old figures already lies 100% inside the final 2.filter GTI; therefore the absence of data in the standard-rejected segments is not a physical lightcurve feature, but a blank produced by plotting after filtering. The uncalibrated all-event − corner in those figures also cannot be interpreted as a FOV/corner SP excess.
Open the corrected selection evidence page →. The new page uses the complete 2.1.filter/*-ori.fits raw time axis, overlays both GTI sets, and shows the archived 2.1 histogram (mean and ±1/2/3σ) separately from the independently parsed 2.filter QDP red selection limits.