I don’t think there is a bug in msi2lmp. I checked dihedral energies on a test molecule using both LAMMPS and Materials Studio, and they agree even though the dihedral energy constants are off by a factor of 4. In LAMMPS, the total dihedral energy is a sum over all dihedral angles, while in some other codes the sum omits equivalences (e.g., multiple dihedral angles with the same C-C bond). So it depends how the energy terms are summed in each MD code.
Also, msi2lmp takes the dihedral parameters out of #torsion_1 cvff. I changed a parameter in that section of cvff.frc and saw the change in the resulting lammps file.
Thanks for helping to resolve this.