Rings in the united atom force field generally do not have impropers (see PHE and TYR as well). That's the way it has always been.....I don't know if it is the best solution or not. The planarity is maintained by the bond and angle terms.
When the Weiner et al. (1984,1986) force field was developed, improper torsions were designated for specific sp2 sites, as well as for united atom CH groups - sp3 carbons with one implicit hydrogen. In the sp2 case, the purpose is to make it harder to move out of plane, while for sp3, the improper was used to prevent inversion of the stereochemistry.
However, not every sp2 atom had an improper applied in the PREP file. When Leap was developed, a decision was made to have the program automatically flag each sp2 atom and check if an applicable improper torsion parameter could be found in the force field - similar to the method used to apply other parameters such as bonds, except that "missing parameters" are tolerated. To prevent impropers with "wild card" atoms from being applied inconsistently with the original "Prep version" of the force field, all such terms were replaced by specific impropers in a special "Leap version" of parm91.dat, parm91X.dat. However, it was impossible to force exact equivalence, since in the "Prep version," impropers had not been applied in a topologically consistent way to the TRP, PHE and TYR residues. We settled for allowing Leap to apply the reported impropers to PHE and TYR (which makes these residues more consistent with TRP), since (1) the new force field was in preparation and contains various adjustments to torsions and impropers that go beyond this issue, (2) we did not want to revise something that had been in the literature for a long time, and (3) except for highly distorted structures, which would be cleaned up by minimization and dynamics anyway, the properties of simulations should be very little affected (as we pointed out in Cornell et al. (1995), part of the floppiness that has been observed in Weiner et al. aromatic residues is due to a too small proper torsion potential for a aromatic C-C bond). Nonetheless, persons publishing papers using the Weiner et al. force field with leap topology generation should explicitly state this in the methods section of their papers.
In developing the Cornell et al. force field, we ensured that all the impropers were specified in a topologically consistent way in the "Prep version," which guarantees consistency with Leap.
Yes, all sp2 atoms should have an improper...so far we have developed those for C=O, N-R and C-R groups, where N is an amide nitrogen and C an sp2 carbon...thus, for example, it would be reasonable to use the same torsional potential for a new residue for any sp2 carbon-exocyclic substituent(C-R) as we have used for all the cases this occurs in the peptide side chains and nucleic acid bases.