To mitigate this problem, if I am recording the spectrum of a target at low altitude (close to 30 deg) I will look for a reference star as close as possible in altitude to my target and then record spectra of the reference star both before and after the set of spectra of the target. I compute response profiles from both sets of reference star spectra and take the mean of these to use in correcting the target spectrum. Although not perfect, this is better than choosing just one of the response profiles. It is instructive to see how much the response profile can change over an hour for a star at low altitude.
David