The tester in this particular case is Peter Acklam. He has been responsive in the past, so I've notified him of the problem.
Peter noticed that the false FAIL reports were specific to perl-5.14.x, and that the problem vanished if EU-MM was updated to version 6.58.
The next release of Math::MPC will therefore specify EU-MM-6.58 as a PREREQ_PM, which should avoid these FAILs.
There's nothing in the Math::MPC distro that actually needs version 6.58 or later, so this change is made solely for the purpose of avoiding a CPAN::Reporter bug. (I hope it doesn't inconvenience someone who is actually wanting to use Math::MPC ... I think that's unlikely.)
The FAIL count because of this single issue has now risen to 42.