My intuition is that you’re not landing on the same SCF state each time. In other words, when Psi tries to find the molecular orbitals, it’s finding an excited set of molecular orbitals instead of the ones you want. This is a common transition metal problem.
Look at your Hartree-Fock/DFT solutions and check how similar they are to each other. If subsequent iterations agree to within the third decimal place, your SCF is fine. If they only agree to the second decimal place, you may have an SCF problem. If they only agree to the first decimal place (or not even that) you definitely have an SCF problem.