I must admit that I do not know how to calculate an objective EV value. I can offer a few things. First, a handful of examples from my most recent test can be found here, including some taken with the camera calculating the exposure with no flash and which can be used to determine the objective lighting conditions from THIS test:
https://drive.google.com/drive/folders/1CPVtXNAfrPp8FI3ovWep5pR09o2l4OSG?usp=drive_link
However this test was a bit better than before. Details on that below.
Second, based on one photo I have from the prior tests with no flash, I think the correct exposure for that situation would have been in the area of 1/160 f4 ISO 3200 +2EV.
The original photos that I started the thread with were taken in an EV equating to 1/160, f2.8, ISO 400, flash at 1/4 -.33 at 200mm, focus distance recorded as 2.02 m.
Now, as to the update on the latest tests ((the ones that I posted files from above):
Today I installed the 2.01 firmware and when I sat down to get a few samples of this problem I noticed that the focus was behaving more consistently. I took around 100 shots, about 80% with flash, mixing subject detect and single point. My initial impression reviewing on camera was that it was behaving much better than it had been. It looked like almost everything was a focus hit.
When I put them into NX Studio for review I found the truth is a bit more mixed than that. On the one hand, I would say that it did behave better. On the other hand, it was still off about half the time or more.
The big difference, and the reason I thought when reviewing on camera that it was much better, is that a much smaller number of the shots are such significant misses that it winds up with the entire person out of focus. Most are misses where careful examination will reveal that the wrong eye is in focus but there's really just a slight difference between the two eyes so that when viewed at a more normal size it might be missed. Some are misses when the wrong eye is in tack sharp focus, and in these cases the other eye is totally out of focus. A handful missed badly so that the ear is in focus or something else altogether in front of or behind the person.
Of course, this probably means that critical focus is still being missed. This is because when the focus is where it should be, the depth of field is such that the in focus eye is sharp while the other eye is totally out of focus. When the focus misses more significantly, the wrong eye is in sharp focus and the eye that had the AF point on it is totally out of focus. Therefore, when both eyes look okay with one slightly better, it probably means that the focus has actually missed somewhere in the middle.
This also agrees with some of Horshack's findings. One, it wasn't really missing (at least not tonight; it did before) when the person was square to the camera. Two, when it did miss it most of the time missed by focusing on the wrong eye.
Now this was just one quick test and I may try it again and find it working again like it did before. However, if the results of this test are accurate - and they do agree with Horshack's results so far - then it would seem to indicate that what is causing the out of focus effect may be that the system is somehow "behind scenes" just plain focusing on the other eye. This would also perhaps explain why the system seems to perform better with single point (though I didn't get much success with that at my event last night): because the actual focusing is working correctly, but the subject detection is for some reason sometimes telling the system to focus on a different eye from the one it reports.