Hello Michael
This particular diagram is used to show the logic behind the creation of vSwitches. In this particular case, VM1 cannot communicate with VM3. You need some L2 connection (either virtual or physical) between those network segments for connectivity.
If all VMs and both the vSwitches, and the physical switches are on the same subnet, then yes, they will all learn the MAC addresses, just like they would if they were all physical devices involved.
It is for this reason that there are other options you can use to design a datacenter to reduce the number of MAC addresses that must be learned. These include the introduction of routing so that you have smaller subnets, and smaller MAC address tables, or the use of other features such as VXLANs which allow you to overcome limitations involving STP, limited number of VLANs, and large MAC address tables.
I hope this has been helpful!
Laz