I am sure that the alignment is correct because the target and template
proteins are the same protein.
If the alignment is correct (i.e. the restraints generated from the
alignment are correct) then the only other possibility is that you have
user-defined restraints that are conflicting with the
alignment-generated restraints. You can try modeling without them to see
if that fixes things. If you can't figure it out, I can certainly look
at your input files.