You get different information with the two approaches. As the quote says, you get the richest data if you use both. If you scan Azure VM hosts using host credentials, you will get Host nodes in Discovery, the same way as if you scanned on-premise virtual machines. You won't get VirtualMachine nodes or information about the cloud service they belong to, or any other information about non-host cloud resources you are using.
With just access to the Azure API, you do get information about VirtualMachines, all the other discovered cloud resources (like load balancers, network configuration, and so on), but you don't Host nodes and the deep details about the software running on them.
If you have both kinds of access, you get all the information and it is all linked together as you'd expect.
Ahh that makes sense, for some reason I misunderstood it Thanks!