vCloud Director should only create the System vDC in a few scenarios. One is if you are using Elastic Pay-As-You-Go pools in vCloud Director 1.5.x.
The provisioning will fail so long as vCloud Director thinks the Resource Pool doesn't exist.
At any point in time, was DRS disabled or the Resource Pools deleted? If this is true, then there is a slight disconnect in vCloud.
The following suggestion is somewhat risky, but might work just fine.
1. Delete the System resource pool that is there
2. force the provisioning to a different datastore
3. this should re-create the system vdc resource pool
4. manually move the two shadow VMs back into the System vDC resource pool