Can you tell what Stop state the cores are put into before and after your command? Technically the cores can be power-gated hence you can make larger-core-count modules behave similarly to the small-core-count variants. The frequencies won't be identical, as the boost tables may adjust frequencies commensurate with extra regulator headroom but it won't line up identically to lower-core-count-tuned modules. Higher frequency could affect your power readings, but taking cores offline should reduce power far faster than raising voltage/frequency, especially with 16->2 cores.
I'm not sure how well Linux supports all of the power management features of POWER9, though. Are all cores active when at idle, or has it already power-gated many such that formally disabling them doesn't actually change their state?
May be worth posting your power/current readings if you have them. VDD current is what matters, if you can separate that out.
Regards, Eric