I have a certain sympathy for the idea - for example, it makes a certain amount of sense, when a person's date of birth is in dispute, to get a copy of their birth certificate. However, this isn't foolproof - I remember someone saying that their birth certificate said "February 30". A good researcher would have taken the primary source material, and put it in context. My grandfather's birth certificate said November 11, but he always said that he was actually born November 1, but that there was a fine for late registration, so his parents gave a later date. If he was notable enough to have a Wikipedia article (he isn't) I could have consulted the original archival material, and said: born [[November 11]], [[1906]]. I could not have used his story to say: born [[November 1]], despite what his birth certificate says (as that's OR). A real historian would have taken the archival material and the anecdote and cross-checked against the story of their being a fine for late registration, and maybe determined from other sources whether such practised were commonplace at the time...and then used that to decide the plausibility of the date of birth (and published it, of course).
Unpublished primary sources can only be valid sources about the content they contain ("his birth certificate says..."). In the hands of an amateur, they really can't be taken to say anything more, we can't evaluate how good the source is, we can't determine how much weight to give to one source as opposed to another...