“I would have thought that, as long as it’s clear what you do (I uncheck the IR box on the dbase entry, and insert appropriate comments in the header) its OK to use an uncorrected spectrum?”
I think we first need to find out why your instrument response is not consistent and independent of the reference star, which it should be if nothing is changing. Otherwise you could be seeing similar variations in your target spectra.
Once that is sorted you can chose not to make an instrument response correction but in my view the spectrum should then be rectified and the flag set in the header rather than post a non calibrated spectrum, otherwise you leave the lamp spectrum imprinted on the continuum
Cheers
Robin