iraf.net

Hello Pierre,It is often the case that strong night sky lines do not subtract well
from long slit data. You are probably not doing anything wrong. The
normal path is to sky subtract after transforming the image since this
will align the sky lines with the rows or columns of the image.
However, since the transform step requires interpolating the data there
are times when the subtraction works better before transform; that is,
if the sky lines are reasonable well aligned with the image rows or
columns in the raw data and the sky regions can be taken fairly close
to the object spectrum. So you could try subtracting before doing the
TRANSFORM step.Another possibility is to not do the background step at all in 2D;
especially if you are dealing with stellar spectra. In this case
extract the dispersion corrected object spectrum without background
subtraction and extract a sky region. The task SKYTWEAK can then be
used to make small adjustments in the dispersion and intensity to make
the best sky subtraction you can.Another approach is to mask out the regions as you mentioned.Cheers,
Frank Valdes