We’ve changed the FCIDUMP code between 1.2 and 1.6, so differences are expected. Changes I’m aware of include changes to symmetry handling, frozen orbitals, and whitespace. For a system with degenerate orbitals, it’s also possible that 1.2 and 1.6 converge onto the same solution but with orbitals that differ by a rotation among degenerate orbitals, which could lead to different integrals. This is, however, not going to affect energies of most electronic structure methods, e.g., SCF and FCI.
If you have a specific reason to believe there’s a problem, let us know, and we can investigate.