A nice result. the Ca H, K lines are showing up nicely in the violet. If you mean the broad hump around 3950A, I have seen something similar on occasion and I suspect it may be an instrument response problem. The raw instrument response from flat corrected spectra produced by the ALPY peaks sharply in this region, for my instrument at least and it is in the area where the Balmer lines are crowding together so it is difficult to fit the response curve accurately around this wavelength. ISIS attempts to mitigate this to some extent by internally applying an assumed black body spectrum for the flat lamp which tends to flatten the measured instrument response but I dont know if Demetra does anything similar. Have you tried running your reference star spectrum back through Demetra using the calculated instrument response to double check that it ends up looking like the reference?