You're welcome to claim cherry picking benchmarks, but the following is, I think, very informative:
This compares the fastest retina Macbook (1.3GHz Broadwell Core M), and the fastest Geekbench score I could find (so presumably running at maximum turbo) to an iPhone 6S.
https://browser.primatelabs.com/geekbench3/compare/2602578?baseline=3515465
I think it's remarkable just how close they are across the board; also interesting is how similar the memory scores are. (You can see this both in the Stream scores, and in the various benchmarks that are memory limited, which are the ones where MacBook and iPhone are pretty much identical; eg the FFTs or Sharpen and Blur.)
This is interesting because in earlier Apple chips the uncore (ie the caches and memory subsystem) were notably inferior to Apple, and the usual doubters were insisting that, OK, Apple can make a quality core but they don't have the years of expertise to make a quality memory subsystem.
(I expect the next argument to be "OK, Apple can make a good memory subsystem, but they don't have the years of experience to make a decent multi-core system that shares the cache well between multiple cores, offers fast atomics and locks, and everything else a multicore system needs for optimal multi-threaded performance...)
I'd say this shows that A9 is definitely comparable to Broadwell-Y; when A9X comes out, I expect we'll see that comparable to the best Broadwell-U's in the highest end MacBook Air.