Granted, phase detection is wonky. That would explain the vertical spread of the data points, but not the horizontal spread.
According to my independent analysis, there can be an effect from hysteresis, IOW depending on the focus approach direction (closer or farther focus position departure point).
I'm not informed about the Focal Pro specifics for each lens/camera body combination, so I'd have to guess about what's actually being reported. It's a pity they didn't choose quantifiable units that make sense, like I attempted in the free
tool I made available here
. Granted, it takes more work than an automated sequence, but isn't the usability
of the outcome more important? My blur metric can also be utilized in most software to directly improve the post-processing quality.
I followed the instructions to the letter. It doesn't make sense why the maximum quality focus that the software/camera can obtain for AFMA +0 would the same as for AFMA +1, +2, +7, +8, +10, +11, +13, and +17 with or without the shotgun phase detect effect. Logic dictates that the maximum achievable quality focus should be parabolic. My results show a practically straight line.
Just guessing, but it seems that the vertical axis, IOW the precision is not good enough to pinpoint the actual optimum (mind you, there may be good reasons, which I don't know, for that).
What I do know is that in principle it's possible to pin-point the actual optimal focus position quite accurately.
For comparison, see below my chart for the lens at 135mm. The same problem. The horizontal spread near peak focus is to broad and too flat.
Similar issue, although it seems not as indecisive as for your other lens.