This was very interesting. I'm looking forward to the follow-up!
On one hand, it's not really surprising that a motherboard with no functional limitations for the relevant use cases and with fewer onboard components should perform as well, if not better than a more fully loaded alternative. After all, more onboard stuff means more work for the system, busier interconnects, and so on.
On the other hand, I wonder if there's a link between the increased power consumption and the increased performance. Might the H110 board somehow be ignoring the power limits of the CPU? Otherwise, shouldn't the power deltas between idle and Prime95 be virtually identical with identical CPUs? Prime95 stresses neither IO, RAM or PCIe, so the list of reasons for increased power consumption outside of the CPU itself is short ...
Skylake ditched the FIVR, right? So, might this board somehow be overvolting the CPU? Not that that should matter for performance unless it's running above stock clocks as well, but I can't make sense of this otherwise.
Did you by any chance log CPU frequencies during benchmark runs?