Rosetta is aggressive about changing the atom names of hydrogens to the ones it wants. The PDB format standard on that has changed at least once and so Rosetta will match one standard and mismatch another.
In all cases, branched sidechains will end up with 4 character hydrogens – H for hydrogen, a letter for the heavyatom counter (like G for gamma), and then the two digits for which hydrogen number index it is on which heavyatom index it is. The difference in standard is whether they are presented as 1HH1 or 11HH.
Rosetta will only be writing to the “wrong column” if the atom numbers get too large (beyond 99999 I guess), and then it will be to the right not the left, so that won’t explain column 12. 0 indexing might explain column 12, a text editor will be using column numbers 1 lower than the PDB standard indexing because programmers can’t count. Your picture didn’t come through so I can’t see what you meant.
The atom names are not made generatively, they’re hardcoded into the parameters file…maybe if you have some very long triply branched noncanonical you could run into trouble where it makes up a weird name? I think there it just assigns them ordinally (like 1H, 2H, … 99H) so you’d need 1000 H atoms to get into the 5th column…at which point something else will have already broken.
.
