Some narrow spike-like peaks in the addition to the normal peaks in the RDF

Hello GROMACS users.
We are experiencing some narrow spike-like peaks in the addition to the normal peaks in the radial distribution function (obtained by instruction gmx rdf in the Gromacs software environment). Simulations are performed for a liquid having the possibility of interatomic H…X bonding. The spikes are suspicious of origination from the intra-atomic interaction. Does anybody have information that indicates the Gromancs algorithm is designed legitimately and purposely to produce such RDF data or experienced such extra spike-like peak as artifice?

Intramolecular interactions can be eliminated by increasing nrexcl to however large a value is necessary to eliminate them, then regenerating a .tpr file for RDF analysis.