I wasn't aware that CHARMM could read a PSF generated with VMD.

The angle parameters for CT1-CT2-CD, CT2-CT1-CD, and CT2-CT2-CD are all the same, so you could probably safely assume transferability, and add CT1-CT1-CD using the same values for Ktheta and THETA. Note that it's possible some DIHEdral terms may be missing once you fix the missing ANGLE term.

The CNEU patch is a relatively recent addition, and was apparently not tested on all 20 amino acids.

Rick Venable
computational chemist