I've been looking into this as well.
When you jump right with -f the f-value of those records again move from positive to negative. And when you jump left from e the same thing occurs with the t in a[t]=an.
They don't match, as in it's not the same amount of jumps required for both of them to end up on 0. But VQC has referred to the excluding middle and meeting in the middle. Maybe it's somehow related to that (I feel like this is just a wild stab).
Correction, I don't mean t I mean literally an (which makes sense since a is decreasing).
I see that the AN values when jumping left with e for the proper record (not 1, c) appear to align with (0, n)'s x's. Can anyone verify? Is it something obvious?
I see there's holes sometimes, or rather cells that doesn't have a corresponding x in (0, n). For example in (e, 5) they will have jumps with AN = 25, 35, 65, etc which doesn't have a corresponding x value in (0, 5). But for those examples it appears to alternate between valid cell and invalid cell in (0, 5).