I have in a test lab a vSphere 5.5 ha/drs cluster created with 2 ESXi hosts and a shared lun. On host 1 I've created a few vm's. After powering down the server to simulate an unexpected shutdown I would expect that the vm's would registered and powered on, on host 2. After waiting for more than 10 minutes the vm's were still powered off and not registered on host2. While I would expect that reserving 50% cpu and 50% memory in the Admision Control Policy would be enough to guarantee that the vm's should restart on the remaining host, the vm's stays powered off. The cluster shows the following warning after power down host 1 "insufficient resources to satisfy configured failover level for vsphere ha". This warning was expected.
What cause this issue? See the ha'drs config screenshot for more info.
insufficient resources to satisfy configured failover level for vsphere ha"
this warning itself is an indication that your Admission control policy won't be effective, because your current running VMs are not allowing your Cluster to reserve what you are expecting. and that's why that warning is appearing.
Please take a quick look into Page 26 and 27 in following guide. This will give you clear idea about how Failover resource reservation based on % works.
After powering down the server to simulate an unexpected shutdown ...
How did you do this? Did you gracefully shutdown the host, or did power off the host, i.e. simulating s.th. like a power failure?
You can disable Admission Control. There is really no point using it in majority of cases.
Answering your question, the reason why VMs do not power on is because Admission Control has to make sure you have enough resources to satisfy failover requirements. When you have only one host, you cannot satisfy any failover requirements, whatever they might be. Because one host is non-redundant by definition.