The NUV ratio between the S/N in the optimal extraction and standard extraction: nuv-out.ps . There seems to be about a 5-10% improvement.
The NUV ratio between the flux level in the optimal extraction and standard extraction: nuv-flx.ps . The ratio is consistent with 1.0.
The FUV ratio between the S/N in the optimal extraction and standard extraction: fuv-out.ps . These values seem inconsistent and it is unclear why there is a significent decrease in S/N around FUV_MAG=20.
The FUV ratio between the flux level in the optimal extraction and standard extraction: fuv-flx.ps . The flux levels are systematically low for faint FUV spectra. Perhaps this is due to noise (but then why the small error bars?), or a bad shift between the actual position of the spectra and the model profile, or a bad profile.
Recommend: try using NUV model with FUV data, and try shifting the model in the NUV extraction to simulate this effect.