Magnus said
“So it might be that the different profiles are captured at times when the skyglow was significantly different. I need to check this more. If it makes sense.”
Yes accurate sky background subtraction is important particularly at this time of year for me too as any remaining offset will change the resulting response curve It will work ok on the original reference star but then give to wrong result on other stars due to the difference in background offset relative to the brightness of the target. Chose the spectrum binning zone so it includes the full width of the spectrum and the sky zones so they are close to the spectrum but do not include any of the spectrum. Stretch the image to see this better eg here Nova Vul 2024 in twilight
