It's a matter of how vR Ops is representing the reservation. vR Ops doesn't say the VM is demanding the memory, it is saying it's demanding the memory on the HOST upon which that VM is running. That "demanded" memory is fully reserved to that VM, so it may as well be marked as effectively used and can't be shared with any other VMs. This is correct, for the most part, because per the capacity utilization model it could be "allocation" or "demand". Surely allocation isn't what we want, since that is just what is configured pretty much, so that leaves "demanded" on this little workload graph.
The only problem, from what I can tell, is the alert you're getting that you don't want. That makes complete sense - I wouldn't want to get bothered either if I get alerts just for setting reservations. Would you mind attaching a screenshot of the active alert with the symptom chart expanded? I'd like to see what alert definition is catching the scenario and what symptoms are triggering it.
If you find this or any other answer useful please mark the answer as correct or helpful.
