prediction error using alphafold multimer

Hi,

I'm designing antibody for my protein using RFdiffusion and ProteinMPNN.
subsequently, I'm predicting the structure of the sequence obtained from proteinMPNN using colabfold alphafold multimer.

I have a question about the structure obtained from colabfold alphafold multimer.

first, my protein is dimer. so, I predict that, designed antibody interacts only one protomer of dimer.
however, designed antibody seems to combine (interact) center of the dimer.
I think the antibody is slightly shifted.

Is there any solution?
What can I try?

Review the reported RFdiffusion → ProteinMPNN → ColabFold AlphaFold Multimer workflow and reproduce the dimer/antibody prediction described in the issue. A useful resolution would identify whether the central binding geometry is expected or a reproducible prediction problem and document a concrete setting or workflow to try; the issue provides no sequence, notebook, or test.