Why wouldn't you have all 3 servers be a part of the cluster so they can share config and endpoint data?
If you have L2 between DC and DR, you can setup a VIP and allow DC to be the primary server. You can also set DR as the standby-publisher in case DC goes down completely. This way you have the ability to make config changes if needed while DC is down, for example if you do guest registrations, only the publisher can generate the accounts.
If you don't have L2, you won't be able to define a VIP between the DC and DR locations, however, you can still add all 3 servers as RADIUS/TACACS servers to your network devices for redundancy as well.
We have 1 CPPM at DC and 1 CPPM at DR, both on L2 network with VIP. They are all within the same cluster.