I was wondering how you got on with this as I will be looking into probing on a horizontal mill also. Did you get a secondary check done off the machine?
My fear is that even if your probe is accurate enough to allow for backlash (the effect of which would be compounded at 180deg) there is no way it can compensate for other 'movement'.
I run quite a large horizontal and use a macro to reset the work offset to suit B-axis rotation. To put myself at ease I regularly check these kind of features using an indicator in a chuck long enough to poke through both holes, and I find variation in accuracy, sometimes over the course of one shift.
Because the work offset is calculated automatically there is no human factor - the actual centreline of the pallet is moving. This could be down to growth, lean, backlash, whatever.
I can compensate for the error using EXT offset (it can require up to +/- 0.02 to ensure alignment) and I always approach 'unidirectionally' to eliminate the backlash factor. If the tolerance is very tight, and if I have access, I will even turn it 90degrees and double-check 'side of hole' Z-depth with indicator. If that tallies I know my B is indexing accurately.
All these measures merely compensate for the apparent error in one axis. I have no way to compensate if the other axes are out of square.