The forces between atoms are computed in CHARMM w/o regard to the molecule type; for H-bonds, the partial charges and VDW radii play a major role, esp. in the close approach distance.
If you are comparing to another simulation result, you should probably try to use the same distance and angle criteria.
Otherwise, I suggest trying the COOR HBOND defaults; we've done that in a publication to evaluate lipid:water and lipid:sugar H-bonds.
Making sure you've properly accounted for image interactions may be a more important issue, esp. if the peptide reaches the edge of the simulation cell during the simulation.