I have converted over to the latest guider mirror setup and I made some tests today using the internal flat lamp and a pinhole in place of the slit (actually a piece of foil with a slit in it overlaid at right angles over the middle of the original slit.) This effectively simulates an on axis star in the spectrograph, though the effective focal ratio may not be correct.
I first checked that the crossed slits were central in the guider image so the slit and my guider image should be centralised on axis. (Hopefully the guider image coma will be minimum at this position too when I make a test on the sky)
I then recorded flat and calibration lamp spectra at H alpha 1/3 from top,middle and 1/3 from bottom of the camera field (having previously focused the calibration lamp lines for best focus.)
The first thing to note is the instruction to set the spectrum at 1/3 (from top or bottom) is rather meaningless as the actual position will depend on the size of the sensor. (I have an ATIK 314 with 6.45um pixels so moving 1/3 of the field equates to moving ~2.2mm)
I found that there was no change in total flux or spectrum resolution for all three positions. (though because the effective focal ratio of the flat lamp may not match that of the scope, potential vignetting will need to be checked using a real star)
I did indeed find that the thickness of the spectrum increases when moving the spectrum from top1/3 to bottom1/3. In this case approximately doubling in width (FWHM) from ~ 50um to ~100um or 8-16 pixels (The actual size of the pinhole in this direction is no known as it was just scored in the foil) This seems less than you were quoting Kevin, though perhaps you have a larger sensor/smaller pixels ?
I also made the same measurements on the zero order image with similar results, though the height of the zero order image was a bit less than that of the spectrum, (~2/3) possibly due to the anamorphic factor which makes the diffracted image narrower in the dispersion direction but wider at right angles to it.