At any rate, CGenFF was based upon the improved alkane VdW from the ether FF. As for the carbonyl, non of the current CHARMM force fields contains a properly parameterized ketone group. The ketone was parameterized as part of the latest carbohydrate FF, and the same parameters were transfered to CGenFF. As you say, the carbonyl parameters are probably a more significant variation. Therefore, it is advisable to use CGenFF for cyclohexanone. The only alternative would be to use the latest carbohydrate FF, but its in-ring dihedrals are highly optimized for sugars and might not be perfectly transferable to cyclic ketones.