its an expected behavior. Please find the details in following KB.
the different is because an esxi host is treated as a resource pool when you see at cluster level.-SD [vExpert 14-20, VCIX6-NV&DCV,RHCE,AWS-SAA]
There is only 200 MHz reserved in the System Resource Reservation, but I am losing 4 GHz per ESXi host.
I understand that it is expected behavior, but that seems like a lot of resources given up.