November Run noise estimate

When identifying the 413.4 feature described in PIPE2D-1154 I noticed that the noise estimate around this position is likely to be underestimated (and not only at the exact position of the feature, as can be seen on the spectrum below. Noise seems to be about OK in the red arm (even if slightly underestimated), but is definitely not OK in the blue arm. I have checked on the original PFSObject file that there is no unit conversion issue, as the same behavior appears on the original data. This does not seem to be linked to the object magnitude.

The noise is also heavily underestimated on the calibration star spectra, as can be seen below :

The 1D Pipeline estimates a factor about 40 between the estimated noise and the one present in the data. As discussed at the last telecon this might be linked to the magnitude, another factor I was thinking of is the resolved/unresolved nature of the objects?