Protein-peptide modeling

GROMACS version: 2021.2
GROMACS modification: No

I am trying to start MD for a protein-peptide complex with CHARMM force field. When I treat system as a whole protein (peptide is a part of the initial protein), the final geometry looks alright. However, when I treat peptide as a ligand (and obtain the corresponding .itp on SwissParam), the geometry is completely distorted after NVT equlibration. The small peptide goes from alpha-helix to unstructured extended chain.
The problem is that another protein-peptide complex could not be treated as a ‘whole’ system since the peptide has a bridge in structure (attached). So I can only start it as a protein-peptide complex and in this case the geometry gets broken.
What could be a solution for cases like this?