Prior to the vacuum problem the fast raster was set for 18 mm x 8 mm to get a pattern of 5 mm^2 per spot++. This did not agree with the optim model - too wide. Looking at the target RMS values on the linked entry suggests a ratio of 1.34:1 for x:y or 10.7 x 8 mm in "MCC units" to get a square spot.
The fast raster software assumes free propagation from the raster to the target in spite of the fact that there are six quads in the way. For standard optics this assumption isn't too bad. For the present optics, with small spot requested, it's lousy.