Sorry for the delay.
I just tested your calculations and I think the reason maybe that your calculation of Cu atom is wrong.
According to your comments, the last 3 orbitals should be dx2-y2 (inactive) dz2(inactive) and 4s(active).
But actually, the main part of these orbitals share the same d part: dxx, dyy, dzz (and s).
So the 3 orbitals may be mixed since Cu atom is sphere.
In your calculation, if you check the VBSCF orbitals carefully, the last 3 converge VBSCF orbitals are dx2-y2, a mixing of dx2-y2 and s, and something like d0 but on the x axis. The orbitals are not what you want, especially for the active part. And this should be the reason.
Attached is my VBSCF calculation with a special initial guess which provides right ordering and shape of the orbitals.
This reduce the energy of VBSCF AND VBPT2 of Cu. So I get the following BDEs:
VBSCF: 224.5 kJ/mol
VBPT2: 388.5 kJ/mol
I think this result should accord with the experiments well.
You may take the VBSCF orbitals from my calculation as initial guess and run VBPT2.
Best regards,