I wouldn't say "most accurate", but it's at least a legitimate romanization. Here comes the linguistics:
The Chinese character 閻 /yán/ (pronounced [jɛn] in Mandarin) was originally brought to Japanese with the pronunciation gloss ヱン /wen/, pronounced [jeɴ]. Over the centuries, エ /e/ and ヱ /we/ first merged into [je], then reduced to vowel-only [e], leaving [eɴ] for 閻. As stated previously, this is similar to using the spelling <yen> for 圓/円, which is /yuan/ (as in the currency, pronounced [ɥɛn]) in Mandarin; it was glossed as ヱン /wen/ (pronounced [jeɴ]) in Japanese, but merged with /en/ in due time, which accounts for the English spelling <yen> vs. the modern Japanese pronunciation [eɴ].
Meanwhile, ン /n/ is a closed nasal whose position relates to its phonological environment; it defaults to [ɴ] (like English <ng> /ŋ/ but farther back in the throat), but becomes [m] before labial consonants /m, p, b/.* Depending on the romanization system, it is either left as <n> or respelled with <m>, so that 閻魔 (pronounced [emma]) can be written <enma> or <emma> with equal justification.
* It is also [n] before coronal consonants /t, d, n, s, z, tɕ, dʑ, ɺ/, [ŋ] before velar consonants /k, g/, and, if followed by a vowel, it may fuse with its preceding vowel to produce a nasal articulation, or transposed such that 雰囲気 /hun.iki/ is pronounced as [ɸuiŋki].
Thus, <yemma> is a perfectly legitimate romanization, though archaic and idiosyncratic.