The plugin doesn't deal with scaling factors. It was written long ago by Rasmus Birn, who was my grad student at the time, but has graduated and moved on to higher things. Frankly, it is too hard for me or anyone else to try to modify Rasmus's code, simply to save disk space. If you want, you can create temporary float files with 3dcalc, process them, then delete them, as in

3dcalc -a scaled_short_dset+orig -datum float -expr 'a' -prefix float_dset
... process float_dset+orig somehow ...
rm -f float_dset+orig.*

You can also convert a float dataset to scaled shorts with a command like

3dcalc -a float_dset -datum short -fscale -expr 'a' -prefix scaled_dset

Also, all AFNI programs can read .BRIK files that were compressed with gzip. If you want to do this manually, a command like
gzip -1v *.BRIK
will do the job. If you want to do this automatically for output of all datasets, you can set environment variable AFNI_COMPRESSOR to the string GZIP. For more information, see [afni.nimh.nih.gov] , as usual. In my experience, image data compresses about 30% -- much more if the non-brain part has been masked off (set to zero).

bob cox