Well, it doesn't, of course. The working code that you posted essentially disables all lower-to-upper case conversions except for the first three ascii lower-case letters. Here's a version that handles a couple different ranges (warning to potential users: STDOUT includes utf8 wide characters):

#!/usr/bin/perl
use strict;
use warnings;
binmode STDOUT,":utf8";

my $tim = "abcdef \x{ff41}\x{ff42}\x{ff43}\x{ff44}\x{ff45}\x{ff46}";
print "main::uc( $tim ) => ", uc($tim), "\n";

sub ToUpper {
    return <<END;
0061\t0063\t0041
ff41\tff43\tff21
END
}
[download]

I tried to see if I could make different packages with different case mappings, and it didn't work as hoped for -- in fact, it appears that the first package to define the "ToUpper" and other case-relation functions will set the case relations immutably for the rest of the script.

Here's a test, which I tried two different ways, once calling the two package subs in the order shown, then in the other order. The second sub call always gives the same result as the first call (i.e. both calls always use the mapping created by the first call):

#!/usr/bin/perl
use strict;
use warnings;
binmode STDOUT,":utf8";

my $tim = "abcdef \x{ff41}\x{ff42}\x{ff43}\x{ff44}\x{ff45}\x{ff46}";
Foo1Case::foo( $tim );
Foo2Case::foo( $tim );

package Foo1Case;
sub ToUpper {
    return <<END;
0061\t0063\t0041
ff41\tff43\tff21
END
}
sub foo {
    my $str = shift;
    print "foo2::uc( $str ) => ", uc($str), "\n";
}

package Foo2Case;
sub ToUpper {
    return <<END;
0061\t0063\t0044
ff41\tff43\tff24
END
}
sub foo {
    my $str = shift;
    print "foo1::uc( $str ) => ", uc($str), "\n";
}
[download]

I have to admit, I don't see the point of this feature, except to make up some really wicked obfu.

(updated to add readmore tags)