>Do you mean the star disc spreads beyong the slit? For the stars I am looking at this is certainly the case. In this respect (star >diameter vs slit width) my guide images look liek the one Andy posted above.
Images of the star on the slit can be deceiving if the star is over exposed as the overspill looks worse than it is. Best to measure the typical FWHM of the point spread function of the star when correctly exposed and in focus. Matching the slit width to this is a good compromise as most of the light will pass through the slit and you will optimise the resolution. With your setup 23um corresponds to 2.4 arcsec and 35um is 3.6 arcsec. I would not go any wider as you lose too much resolution. You cannot reduce the star size using a focal reducer either as that produces more vignetting in the spectrograph. When you do these sorts of calculations you begin to see how the telescope, local seeing and spectrograph design are intimately linked. The LHIRES is optimised for an 8 inch f10 with 2.5arcsec seeing.
EDIT: Found the calculator I was looking for. For a gaussian shaped star profile, 76% of the light will pass through a FWHM wide slit.