Must be specific to your machine, distro, or perhaps your $LANG setting (wild guess here); running the above on my Ubuntu 9.10/Perl 5.10.0/$LANG=en_US.UTF-8 setup produces not a single "weird character" in 489 entries, either via 'less' or just on the console.

Is there any sort of a common factor to the entries that give you the "weird" output?