Potentials result in pull to C-terminal end of protein?

I was trying to generate N-terminal homodimeric caps with the following code below:

/rds/general/user/dp5117/home/RFdiffusion/scripts/run_inference.py
inference.symmetry="C2"
inference.num_designs=10
inference.output_prefix=${Results}/${pdb_filename}_designs/${pdb_filename}length${length}
'potentials.guiding_potentials=["type:olig_contacts,weight_intra:1.0,weight_inter:0.1"]'
potentials.olig_intra_all=True
potentials.olig_inter_all=True
potentials.guide_scale=2
potentials.guide_decay=quadratic
inference.input_pdb=$pdb_filepath
"contigmap.contigs=[$length/A24-228/0 $length/B24-228/0]" \

However, this resulted in the capping regions expanding outwards to the C-terminal interface, see image below.

Removing the potentials flags rectified this and I ended up with plausible N-terminal interfaces.
Curious to see what the authors may think is going on?

DP

Start with scripts/run_inference.py and trace the supplied symmetry, contigmap, and potentials options, especially olig_contacts and guide_decay. Reproduce the C2 design with and without the potentials flags and compare the resulting N- and C-terminal interfaces shown in the report. Done means identifying why the potentials shift the cap or confirming a configuration or modeling issue and documenting the resolution.