Printing Anomalies

I'm printing small parts with a Longer 3D Orange4K Mono SLA resin printer.  I'm using a fairly cheap resin: ANYCUBIC ABS-Like Resin Pro 2 Gray.  Printing the pictured little part flat (first image) upon the glass/build plate produces excellent reproduction/precision for all surfaces except the bottom surface, which is attached by supports to the build plate and is always messy and unacceptable.  The precision of the holes, top surfaces and vertical surfaces for that orientation is PERFECT through the entire Z range, as you would expect.

My intuition tells me that mounting the part (main body approximately 2mm x 15mm x 6.3mm, two 3mm holes) vertically (second and third images) should print without problems.  With a vertical mounting, printing the first half of either of the two large holes is no problem, because they're supported at all points above for every slice, as the build plate during lifting fights against the 'desire' of the slice to remain attached to the FEP and (hopefully) wins, pulling the slice up with the build plate. Beginning half-way down each hole, there is a slow progression of material back towards the center from either side of the hole, as more shearing/bending strength (or lack thereof) becomes involved instead of primarily bulk elongation/tension strength along the edges of the hole at the current slice in contact with the FEP. But the hole flattens out near the bottom of the hole, and it does so despite the tiny supports sitting in either hole (second image). I would expect the opposite. As the current slice approaches the bottom of either hole, the shearing/bending force upon the previous slice(s) approaches maximum. So, I would expect the void of the hole to be pulled AWAY (downwards) from the center of either hole somewhat and the holes ELONGATED. Not drawn in (upwards) towards the center of either hole and shrunk (as pictured in the final product). So, it remains mysterious why I see the flattening of the hole, unless there is some 'recoil' involved in the winning of the tug-o-war for the slice. Is there an error in my reasoning? Why is that happening? Does anyone understand the physics? Could the problem be the cheap resin I'm using?  Yes, I know that printing with off-angle orientations produces generally better results but with generally less precision and more 'noise' also.  Thanks.