pull request #7 from spawngrid/fix-tuple-call-r16 breaks exat

[jnorthrup]

in http://github.com/gleber/exat/issues/5 I encountered a regression far beyond my erlang-fu to diagnose properly, but I created a brute force regression which went back quite some difference from the mainline master HEAD.

I would very much appreciate knowing how much is sacrificed or what can be cherry picked to improve the coherence of these two packages.

Thanks in advance
Jim