I think this is by design. assuming that in Starwind you enabled clustering when you created the targets, your Hyper-V nodes are what's doing the reservation. In MIcrosoft clustering, only one node at a time can own a disk resource, so you should find that one node has the disk online and the other has it reserved. Is the cluster working OK? If it is, you should be able to quick migrate one or both of your VMs to the other node. If that works successfully, when you look at disk manager you should see that the server that was marking a disk as reserved now has it online and vice versa. Also, are you using one of the iscsi targets for both a VM and your cluster quorum? I wouldn't recommend that.... give the quorum its own iSCSI target. This can be very small; 512MB is plenty.
Hope this helps,