Hmm, not easily. I got the list of names, but the person/group flags seem to mostly be unset. I tried a pass filtering out anything with a comma in the sort-name (other than ", The"), but that isn't used pervasively enough. I extracted some common first-names and filtered those out, too, but the results still look too J-heavy to me, so I don't trust the numbers...

This is easier than it sounds, since MusicBrainz has a huge database of artists across all musical genres, and they also have it pretabulated by first-letter. But harder than it then sounds, because their list includes individual people as well as groups, and personal names have a very different distribution than group names, at least in English. The MB database does have flags for whether an artist is a person or a group, though, so this ought to be possible, if I can figure out how to extract it in bulk...

Ah - but the implication of the task in the first place would seem to be to investigate an inkling that certain letters might be more or less popular as first letters of metal bands' names. And that means that you'd need to compare the metal name list's frequencies with the frequency of initial-letter occurrence in the language as a whole (or, if it's an international survey, collectively among the languages represented in similar proportions to the included metal bands' languages of origin), and then compare that to the first-letter frequency of metal band names. Here's where my suggestion breaks down, in that I'm not a statistician and can't quite describe how this would work...but it would seem a fairly simple task of teasing out significant differences within the metal sample that exceed chance variations resulting from the smaller sample size of the band names.  

So, you know, if anyone's not doing anything, have at it. And then do the same thing for acts in other genres, so we know it's metal-specific and not just band-name specific. (Seems like lots of punk bands whose names begin with "X"...or for two bands, *are* it.)