You're absolutely right. I had, apparently mistakenly, assumed that @_ was a superset of the contents of both @EXPORT_OK and @fail, instead of a subset of either.
Lacking the context of how @_ is being defined, it was an easy assumption to make. Without visibility into the calling code, I'm reticent to hazard another guess.
I suppose it's not oustide the scope of reason that the code may not reflect the author's intent. In a module this pervasive, however, it's not likely that this is the case.
-><- -><- -><- -><- -><-
All things are Perfect
To every last Flaw
And bound in accord
With Eris's Law
- HBT; The Book of Advice, 1:7