More MHZ does not equal better consolidation ratio ?

I am doing an assessment with these two different servers as target system for server consolidation:

An IBM x3690 with 20 cores at 2,400 MHz wich totalize 48,000 MHz
An IBM x3650 with 12 cores at 3,460 MHz witch totalize 41,520 MHz

Everything else is the same on those two templates (Ram, IO, queue...)

I was under the impression that the x3690 would be chosen as better performing target by the CP compaired to the x3650 but this is not the case.  I end up with 3 phantom systems in both scenarios but the processor utilization is higher on the x3690 (lower consolidation ratio).  This does not make sense to me.

Anyone have an idea why ?

Thanks in advance.