I think the details here are important.  A quick look at the Laserline Optics website shows that they offer different Powell lens design options for each fan angle to accommodate a range of beam diameters.  However, they also have “narrow” and “wide” beam options (with different CAD models for each) and it’s not clear which one you get for given beam diameter (or if you have to specify this detail in addition to your beam diameter):

In any event, if you are using the correct CAD model (with the right glass, S-TIH6) for your target Gaussian beam width, then you should be getting reasonably reliable results.  Not sure what your wavelength is (1.08 um?), but I would suggest you contact the company and make sure you are using the proper CAD model for your particular problem.  Also, as Sean mentioned, you want to confirm the CAD model has sufficient resolution for ray tracing (I think it’s probably okay, but that’s another question for the company).

Lastly, I see you are tracing only 5,000 rays, while using a detector with 2000 x 2000 pixels.  I would suggest you use significantly more rays, like 1e06, and reduce the resolution of your detector (to say 500 x 500 or less) so that pixels that are being illuminated will receive enough rays to provide a decent SNR.