” why and how does reddening matter? OK, the actual spectrum does not match what we would expect – but as long as it is consistent as it is, could that not sufficient for a good ref-star? ”
If the spectrum is reliable then yes, reddening should not matter as you have shown here. 0.6 reddening is a lot though (~80% of the light in the V band is being absorbed) and I prefer my reference stars to be as “normal” as possible. If the (B-V) matches that expected for the spectral class as it should do for a star without IS extinction that gives added confidence that all is well. (Reddening of course matters and needs to be low if we are assuming a spectral type and using the Pickles spectrum)
” I use a camera with relatively low dynamic range (8 bits)”
8 bits is not really enough for serious work but are you sure your camera is only 8 bits? the ASI183MM spec says it can run at 12 bits (often stretched to appear as 16 bits in fits images by the software.)
https://www.zwoastro.com/product/asi183/
I don’t have much experience of CMOS cameras but as a starting point I would suggest setting the gain so you use the full 15k well depth, expose so the maximum is around 2/3 of saturation and average many flats (eg 20-30) to beat down the noise at the blue end, not forgetting to subtract bias and dark when [preparing the master flat (The ISIS “masters” tab handles this)
The other thing to watch out with flats is stray light in the spectrograph. I don’t know what the LowSpec is like but with the LHIRES, flats at the blue end are unusable as most of the light comes from stray light, not from the spectrum !
