Thanks for the suggestions & references. I’ve only been able to look at switching off optimised binning so far, but it makes no difference to the results. On the way I checked out switching everything out [no offset, dark or flat field correction, just raw images with no calibration]. I looked at the consistency in pixel location for several absorption lines across all the images and the standard deviations seem reasonable to me.
H.alpha=0.15px; H.beta=0.13px; H.gamma=0.11px; H.delta=0.16px; CaII=0.09px
I will look again at the data to see if the results show any obvious trends with time. However, more disturbing is that I have found what looks like a straylight problem with my set-up. If I turn up brightness & contrast to the maximum in the calculated image means, there is a streak of light running parallel to the spectrum [in all the images, to varying degrees]. It is approx 3% of the peak of the signal and about 50 pixels away. It wouldn’t have been summed in background removal, but not a good find. Am concentrating on this for the moment.