To give a bit of context with this dataset :
We move all cobras behind the dot in the first place and then we move the cobra out of the dot, group by group and take an exposure each time.
So what I did is calculating the flux gradient between visit[i+1] and visit[i] and find the group that was moved matching the maximum gradient, that's the "mostLikely" column, in a perfect world, actualGroup always match mostLikely.
the sig1 and sig2 columns gives you a sort of confidence value, with

sig1 = gradient[actualGroup]/gradient[actualGroup-1]
sig2 = gradient[actualGroup]/gradient[actualGroup+1]

if the mapping is correct, their value is above >1, the higher the ratio, the more confident you are.

Here, i'm showing only the fibers with sigma<10.
there are few fibers with sigma slighty above 1 and only one mismatch.
The only mismatch, is basically in the noise, with less that 1% variation.

I've also added the final result from the roaching prior to the fiberIdentification sequence where cobra were not successfully hidden.
As one would expect, they match perfectly, the low sigma in the fiberIdentification dataset.

So I think, we can conclude that there are no fiberMapping issue in sm1 or sm3.