I strongly disagree. The only cites should be good cites. Bad ones are misleading and wasteful. A good cite is easy enough to make: one that accurately describes the cited material and relates it to the subjectmatter.
Maybe I need to state what I meant better; citations should be good BUT one should cite references to one's sources even when those sources are not necessarily the best.
I think there are two separate things being discussed here, which are being confused together (by me at least, and thus I assume probably others).
One, the quality of the citation. This is what you described above. In other words, it identifies the source accurately, gives enough information to find the needed information easily, and documents its relation to the subject matter truthfully and accurately. This has nothing to do with the actual nature or quality of the source material. A high quality of citation is essential.
Two, the quality of the actual source material. This is a quite different matter, and contains much more subjectivity and scope for argument (not to say there isn't any in the previous). One can have good cites for bad sources, and bad cites for good sources.
What I really meant in the previous post Tony replied to is that one should always give good cites for one's sources -- even sources not of the highest quality, accuracy etc. A good cite for a bad source beats a bad cite for a bad source, or no cite at all. At least the article is properly referencing and characterising its sources, and is therefore much more transparent.
Bad cites are an insiduous poison in that they claim authority but lack it, and are deceptive. Bad sources are easier to fix or at least identify.
Sometimes one is forced to rely on poor sources while an article is still in development, while one is searching for more authoritative reference material. On some subjects there are no good sources at all; all are unreliable or incomplete. In these cases, it's best for the conscientous contributor to use what sources are available, cite them meticulously, mention the problems in the article if it can be done in a NPOV way, and certainly document the problem on the talk page for future editors to read.
I hope this clarifies things.
-Matt