There's only so much a compiler can do. We're talking about a few percent here, and even that only applies to high level code. Also, while MULTI does produce fast binaries, it tends to favor size and stability/ security over speed. In the end, it all comes down to the developers, and they have years of experience with PS3 and 360, but only a few months on Wii U.
In fact, we had similar issues last gen: Code written for Xbox often performed worse when ported to Xbox360 even though Xenon is a lot faster then some old Celeron. There've also been issues when Intel switched from insanely high clocked Pentium4 chips to much lower clocked Core CPUs - some code simply performed much better on P4 even though we all know Core is far superior overall. And those were both x86 chips, just like Wii U and Xbox360 both use PPCs, so that doesn't necessarily tell us much, either.