This is my justification for deletionism.
Wikipedia's social processes work best when Wikipedia is a tertiary source. It is not a good repository for primary source information (although WikiSource can work), because we don't have the type of processes built in at preserving or verifying the meta-data of primary source material. It is not a good repository of secondary articles (that is, original research) about primary source material. Secondary articles belong in a place that has fact-checking as part of its mission somehow: a newspaper, a peer-reviewed journal, etc. Wikipedia is a tertiary source: we summarize and synthesize all the secondary source material into a coherent whole.
As an example: if you are a researcher and have found a new letter between George Washington and Benedict Arnold in somebody's basement, you should get it checked out by other people competent at verifying it. Wikipedia would never simply accept you typing out the contents or scanning in an image. It is not competent at verifying such a letter. If you are a researcher and have a new theory about George Washington, you should publish that theory in a scholarly journal of historical research, vetted by other historians; Wikipedia is not the place to argue for that new theory. We're simply not competent at evaluating it. Wikipedia can come along afterward and describe the debate in the peer-reviewed journal, but it's not good at the peer review in the first place.
When the subject is not George Washington, but rather Joe's garage band, this breaks down, because it's impossible to be a tertiary source about something which there are no secondary source materials. What if it turns out there is no Joe's garage band? Joe doesn't even own a guitar? There is no Joe? Even a really bad music magazine would probably suss that out, but Wikipedia wouldn't, because we don't (and pretty much can't) have a good process for editing secondary articles like that. And if it's pretty much impossible for anybody at Wikipedia except Joe (if he even exists) to verify it in any way, then the Wikipedia processes can't work on it.
Wikipedia has to draw the line somewhere (do we want 366 articles on "What Joe ate for breakfast on January 1, 2008", "...January 2, 2008"?) I think a good place to draw it is at our core competence of being a tertiary source.