Thank you for the clarification. I absolutely agree on the beam becoming elliptical away from the centre.
$400 for an f-Theta lens in the UV with scanning range of 125x125mm and working distance below 300mm seems really cheap! Though still making a difference on a $3k machine.
I used a telecentric f-Theta lens in a laser machining system I built, hence the thought about focusing, and I think that this would be the only kind that could solve the spot distortion. With a non telecentric f-Theta lens the beam hits the target still at an angle and, while solving the focal plane distortion introduced by a normal focusing lens and making the scanning displacement proportional to the incidence angle theta, it won’t solve the beam profile distortion. The one drawback of telecentric f-Theta lenses is that they are not able to scan on areas wider than the lens diameter itself, meaning you would need one larger than the buildplate… kinda prohibitive I guess.
The document you posted is very interesting, thank you. My french is quite poor, so I might have missed something, but it is interesting to notice that the offset measurements average were negative (less than the designed value). Sort of giving some insight on the resin shrinkage. No mention is done unfortunately on whether the measurement were done before or after full cure.
Very interesting the ANOVA on the measurements! The F values are not very high in general, quite low for the Z measurements, but definitely show some correlation. I’d be curious to see the raw data to understand if the off-centre position introduced en elongation (due to the stretching of the spot) or just inconsistency. If it was the elongation I would expect a more asymmetrical distribution. In the latter case, the galvo or even the mirror could have an impact. Or maybe the effect of the longer optical path in the vat, incidence angle at the interfaces…
Good someone asked himself the question, but more than a secondary measurement on the printed objects, in my opinion the final answer is into measuring the beam profile directly at different locations and mapping the scanned surface to understand the real cause. Anyone at Formlabs reading and willing to spend some time for the characterisation? 
Thank you again
