Actually, I'm remembering more about this, I've been away too long....
The way it's supposed to work is this. In a normal synced recording, only the beginning of the Record is truly synced to a host pulse. The ending of the recording is not technically synced to a pulse, instead the ending is rounded off so that the loop is exactly one multiple of the sync "unit", typically a bar. This helps correct for jitter in the sync pulses, For MIDI especially, the ending clock may not actually be exactly where it should be because clocks tend to wobble around a bit. It is less of an issue for host sync, but it can happen.
Most of the time the difference isn't noticeable, recording starts on a bar and it then ends after an exact number of bars, if the host pulses aren't jittering then it will still happen to coincide with the ending pulse, but if there is an error the difference is so slight you won't notice.
Where it gets weird is NextLoop+Record which for EDP users is expected to begin recording immediately even if you are not at that moment on a sync pulse. The ending of the loop is however rounded to be an exact multiple of the sync unit so you should get a loop that is the right size even though recording didn't start exactly on a bar. This is the second of the three approaches outlined in my previous post.
Going back to your example, once the recording is in progress you should see cycles added whenever recording fills a snc unit, a beat or bar if you're using host sync. If your track quantization settings were such that the quantized switch happened exactly on a host bar, then the effect should be the same as if you were doing true host sync without NextLoop. But if Switch Quantize is less than a bar, then it may start early but round off to a bar.
I'll run some tests to make sure that's working.