Strange results with m06-2x...

I checked the orbital energies from the NWChem M06-2x calculation and there was no "hole" such that an unoccupied orbital was lower in energy than an occupied one. I've seen that happen in these calculations before, and they've never resulted in a 54 Hartree difference.

I also did as you suggested and did a HF calculation to get a better set of guess orbitals. The total energy for this came out to -763.057301433098 Hartrees, which is nearly identical to the G09 results of -763.057667241 Hartrees. However, using these orbitals as a guess for the M06-2X calculation led to an energy of -713.1746899070 Hartrees after the first SCF iteration. This is very close to the energy of the initial SCF calculation with M06-2x where I didn't use HF orbitals as a guess: -713.17469465. This once again leads me to believe that the issue is with the implementation of the M06-2x functional.