Your Cfactorial case didn't make real use of a closure. You should take out the "my" in the sub and do

$iClosure = shift;
[download]

Your benchmark is not fair to Closure because you are repeatedly creating subroutines. The following would be better:

our $iGlobal;
our $closure_sub = Cfactorial();
timethese (1000000,
{
'Recursion' => 'Rfactorial(30)',
'Global' => 'Gfactorial(30)',
'Lexical' => 'Lfactorial(30)',
'Closure' => '&{$closure_sub}(30)',
}
);
[download]

When I do this I find that the times for Closure, Global, and Lexical are virtually identical.