When a medicinal chemist starts digging around through the literature for help on some chemistry, there are several levels of results. The most welcome are recent papers that solve the exact problem you’re looking for, of course. We’re not in business to come up with new reactions (unless we have to,) so if someone else has done the work, that’s great.
Almost as good are similar reactions from the reliable literature. Different people have different borders drawn for that territory, but everyone would include solid publications like the Journal of Organic Chemistry, particularly if it’s a full paper. Interestingly, that’s because JOC is generally not the place (these days) for the hottest work to appear, which takes that out of the equation. As has been proven several times over the last few years, journals can stumble when they try to publish stuff that’s a bit too avant-garde: some of that work is so cutting-edge that it hasn’t even been done yet.
The communications-only journals vary widely in quality, so they’re generally a step down from the better full-paper publications. One big reason is that the communications often don’t include much in the way of full experimental details. One way to tell how useful a journal is would be to measure how surprised you are when you can’t repeat chemistry from it. If you can’t get some reaction from Tetrahedron Letters to work, you just say “Oh, well”, whereas a dud reaction from a long JOC paper gives you more of a feeling of betrayal.
Then there are patents. When I was a grad student or post-doc, I tended to just ignore patent references, but that was partly because I couldn’t get access to them very easily. Now there’s less of an excuse, and anyone who bypasses them is missing out on a lot of useful preparations. They don’t always work quite as wonderfully as advertised, but there are a lot of very interesting intermediate compounds that are described nowhere else. And when the patent goes on to prepare seventy-five enabled compounds from said intermediate, you can be reasonably sure that you’ll be able to make enough for your own needs.
But there are patents, and then there are patents. If there’s no spectroscopic data associated with a compound, you’d better step lightly. Similarly, there’s journal literature and there’s. . .well, there are an awful lot of journals out there. And some of ‘em are, in fact, awful. SciFinder and other such tools are perfectly capable of tossing out references, in the same list as everything else, to the Bulletin of Some Obscure Country’s Academy of Sciences, 1953, communicated from Unpronounceable University in Everyone Leaves Province. You’re on your own if you track these things down, and good luck to you.
In a category all its own is the Soviet-era Russian literature. There are a large number of compounds (particularly heterocycles) that are described nowhere else, and a wide range of these things are available in English translation. But (as with patents) you have to be careful. Some of this material is really worthwhile and unique, and some of it is. . .well, my theory has always been that people in the Soviet era were willing to do a lot to remain "academicians", considering what some of the other options were like. Can't say I blame them, either, but it means that the more obscure Communist-era references need to be approached cautiously. If you're depending on a reference from J. Siberian Oil Chemist's Soc., (a real journal, at one point), then you may need to start looking for some backup.
Update: On the other hand, here's a new mathematics journal that's made up, explicitly, of papers that have been rejected by peer review. Each includes a summary of why it was rejected, and why the original author thinks it's still important anyway. . .thanks to Marginal Revolution for the tip.