UT theoretical chemistry code forum

I'm doing hydrogen recombination from a surface. So the two leaving H's are bound [at the surface] in the initial state... and for the final state those two H's have beomce an H2 molecule about 5 angstroms above the surface. I have 10 images. Here's the thing I can't figure out: when I watch, in Jmol, the 12 frames of the movie created with nebmovie.pl..... I notice that the tenth image "moved backward"... by that I mean that the ninth image was actually closer to the final state (further from the surface) than the tenth image. How could that happen. Obviously the "nebmovie.pl 0" picture based on POSCARs will not show this "backward move".... so what can happen during a run that can cause this? I'm just thinking in nature it would be shocking to see an H2 change direction like this without interacting with anything (obviously)... so what about the NEB run could cause this?
Thanks!!
(I will admit that this run should have been done with spin polarizatoin on but I did a non-spin polarized run to just try a quick first pass, but nevermind... I don't think that would be related to what I'm asking about)

Is it possible that you endpoint is not fully minimized, or calculated with different settings from the rest of the band? In either case, the moving images may relax further than your endponts -- for example, by moving further away from the surface -- and then the band has to connect to your final state and so the movie would show the H2 turning around.

Just one idea. If this turns out to be a problem, we can take a look at the calculation if you upload it.