Peptide moving out of the box during peptide-protein MD

GROMACS version: 2023
Dear all,
I have done the MD simulations of protein-peptide systems (total 3 MD simulations). After MD simulations, I wrote the trajectory by using the following command:
gmx trjconv -f abc.xtc -s abc.tpr -o abc_pbc.xtc -pbc mol -ur compact -center -n chain.ndx
selected protein for centering and wrote the protein-peptide as an output of trajectory.
After that I did the binding analysis using gmxMMPBSA, I found the negative delta G but when I generated the pdb at the end of the simulation I found that its periodic image was coming into effect on visualizing in VMD. I have also calculated the RMSD and it is coming around 0.9 nm. But when I look at the decomposition results of gmxMMPBSA the residues involved in interactions are different than the residues involved in interaction in end pdb (from ligplot).
But in other MDs I have written the trajectory in the same way and in those cases peptide is not going out of the box, also the RMSD is much less around 0.22 nm and the binding energy is approximately similar to the above one.
Any comments or suggestions will be highly appreciated.

You likely calculated these for the system without fixing PBC. For multiple peptide simulation systems, I always use the multiple PBC fixing method to ensure all peptides are inside the box. Further analysis was done only after ensuring that the PBC is fixed.

gmx trjconv -f abc.xtc -s abc.tpr -o abc_pbc.xtc -pbc mol -ur compact -center -n chain.ndx

when you visualise the trajectory you able to see the pbc effect or it is only at the end frame?

After 100 ns, the peptide goes out of the box and its periodic images come into play after 100 ns and it remains there till the end of the simulations.
I have applied the same command to other peptide-protein systems, in those cases, PBC is fixed but in some cases, I see this problem. I have applied different pbc commands but am still not able to fix this.