Parke Davis stuck the keto group in there to shorten the half life by a considerable amount. It gives tha body a metabolic tab to start manipulating the molecule with an eye on getting it out of the body. PCP doesn't have that long a half life, but can be 'recycled' within the body, diffusing from the bladder back into the bloodstream, making the drug hang around longer than it's wanted. Stick a nice polar tag like a keto group somewhere and not only does the body have loads more ways of being able to metabolize the drug, but it reduces the fat solubility, so it can't re-enter the bloodstream from the bladder as well.
What has to be understood is that the shape the PCP molecule takes up in vivo, makes it a difficult metabolic target, which is why ketamine was developed - in fast (retains fat solubility, sort of), out fast (the keto group makes it easier to excrete) - once a drug is stopped being administered, you (from a medicinal/pharma point of view) want traces of it out of the body a.s.a.p. In terms of that, a hydroxyl group would be better because it allows conjugation with things, but in this case, a hydroxyl caused the potency to drop off the scale.
Basically, for other dissociative effects, play around with the aryl group (such as swap it for thiophene/other aromatic heterocycles or add ring substituents such as 2-chloro or 3-methoxy) or the amine group (sec or tert amine, heterocyclic/PCP or aliphatic/PCE), but playing around with the cyclohexyl ring either abolishes activity or leads to much more potent opioids, moving away from being a dissociative and potentially changing the whole nature of the final activity of the compound. BTCP was a case of modifying it so the DRI activity remained, but everything else dropped off (unfortunately it picked some new activities, like hepatotoxicity)