I would guess not, because Salmon has a specific error correction model for short reads and one for long reads, so putting them both in the same file would mess with that.

I, too, have short- and long-read data from the same samples and here is how I approached analyzing them with Salmon:

I used FMLRC2 to correct the long reads with the short reads, improving the per-base accuracy of the long-read data. Then I mapped the corrected long reads with minimap2 and fed the mapped files to Salmon using the --ont flag.

If you don't want to correct with FMLRC2, and would rather use both datasets for quantification at the same time, you can use StringTie, which has a mixed reads assembly mode (--mix), that takes both long- and short-read files as input to calculate gene abundances, but this may or may not be compatible with downstream tools, depending on what you want to do with the data after (For example, I used Salmon because this approach was not compatible with IsoformSwitchAnalyzeR).