Non physical Forces while interacting with wall/gran/region

Dear LAMMPS users,

I have to signal a problem with conservation of linear momentum in a granular simulation.

My simulation is very easy:

I have 2600 granular spheres poured on a vibrating plate that is modeled as a wall with the wall/gran/region command.

The region in initially at rest and after 100000 steps it starts to move with a sinusoidal z-velocity.

The interaction with the wall acts just in the z direction because i have turned off the tangential component in the wall with the wall/gran/region command. Despite this, the x and y component of the force acting on the center of mass are almost zero but with some huge spikes... Those spikes clearly affects the linear momentum conservation in the x-y directions.

The problem is present both in parallel and serial runs.

I attach to this message the source code of the script that gives the error. The log file contains all the observable mentioned, do you have any idea for this strange behaviour?

Thank you in advance

Your script is complex. I suggest you visualize your system to make sure
the dynamics look like what you expect. What are your boundaires in the x,y directions?
Do they not contribute forces to the particles? Also, if the bottom wall is just
a flat plate, fix wall/gran/region is probably overkill. Just try fix wall/gran.