Your reasoning is solid. When at all possible, I advise to go with a plan to re-create the VMFS5 datastores from scratch. I've written on reasons why here:
Regarding the large VMs: since they most likely will not be migrating around in the future, and you basically have the "one big VM" design on them, I don't see a problem with doing a straight upgrade. Even the smaller sub-block size won't help you out much, since there's only a single VM worth of non VMDK files on them to fuss with.