Hello Roger,
in short: 't2x' converts the coordinates from the gradient file, 't2x -c' the coordinates from the coord file.
The reason that this differs is because the optimiser (statpt by default) checks whether the convergence criteria are meet and then writes out one last geometry. In order to have orbitals that match these coordinates, one last SCF run is done.
If you run 'sdg energy' and 'grep cy gradient', you will notice that there is one last energy for which is no gradient. This is the energy that belongs to the coordinates in the coord file.
If you need the most right energy (say for a reaction calculation), use the one from 't2x -c'.
I hope that clear this up.
Arnim