LJPME should not accept VDW modifiers it does not support
An .mdp including the fragment
vdwtype = pme vdw-modifier = force-switch
is accepted, produces a .tpr that is different from that with vdw-modifer = potential-shift, and IMO is mal-formed. I'm not sure what it computes in practice.
IIRC the only supported modifiers with the Verlet scheme are potential-shift and (maybe) none, so grompp should only accept supported combinations (ie probably not potential-switch, either). We should not just ignore a modifier that can't be implemented for the vdwtype, because we don't know whether the user's error was in choosing the vdwtype or its modifier.
Check this for electrostatic PME, also.