Hi there,

Me again cheers.

As per my understanding after a few readings (I admit, I started from Wikipedia and then did some more research), use of chōonpu in hiragana is extremely rare (e.g. in some manga to add emphasis).
Hence, it not being written down in src/hepburn.js in hiragana trigrams seems normal.

Moreover, Ō seems like it can be transcribed either as OU or OO in hiragana.
E.g. TŌKYŌ is actually TOUKYOU, whereas ŌZEKI is OOZEKI.

toHiragana("CHŌ")
// could be either
// ちょう
// or
// ちょお

That being said, the lib could probably take an extra step and transcribe some valid romaji when there is no ambiguity (i.e. A, E, I and U)
E.g.:

toHiragana("CHĀ")
// ちゃあ

On the other hand, repeated vowels in hiragana cannot safely be transcribed in romaji using the macron (eh) diacritic « ¯ ».
E.g.

fromKana("からあげ")
// KARAAGE
// 唐(から)揚(あ)げ
// KARĀGE would be wrong here

So, transcribing long vowels from romaji to hiragana and back could yield a different romaji, breaking symetrical transcription.

WDYT?

Cheers.