This sounds pretty straightforward. You have the list of datastores, now you want to aggregate the throughput on it. You got the Custom group in the Custom UI populated with the Datastores, now just create a Supermetric, apply it to the Group, sumN(xxxx,1), with the storage stat you need, and it'll roll up to an aggregate (which is what you want).
If I understood you correctly.. You shouldn't need to go back up to the VM RK level, as you can get all of the Disk I/O stats you mentioned from the Datastore RKs. Remember, if you start putting SM on the Datastore to pull Disk I/O stats DOWN from VM to Datastore, you're going to pull I/O that could potentially be coming from not only different VMDKs on different storage, but also cause a situation where you're double-dipping statistics of VMs that have disks that span multiple EMC datastores. Basically - don't do what I just described in this paragraph if that's what you were planning.
If you want granular.. that's not a big deal. Just create dashboards that populate widgets and filter based on the Custom group tag, as it'll filter the Datastores in the widgets to those in your group. Do Heatmap and Top-Ns to find your objects of interest. If you want to see what VMs are running on datastores, just use a resource widget, viewing the parent, maybe a special column with VM disk space, with an interaction and you'll get you're association of what VMs are on which datastore.
With all that I just mentioned, the customer should be happy.