ゔ and ヴ seem not to be included in kana conversion, which causes e.g.
the output of
romaji2hiragana("virama")
to contain katakana. I suspect this is due to the lines
$str =~ tr/ァ-ン/ぁ-ん/;
$str =~ tr/ぁ-ん/ァ-ン/;
in katakana2hiragana and hiragana2katakana, respectively, as the
syllabic n in both sets has a codepoint value one less than the
corresponding vu syllable. Rare as ゔ may be in actual usage, it would
be useful for round-trip compatibility between the two syllabaries.