Even if there are millions of distinct plant chemistries out there, a small number (say, 100) will account for 90% of the herbal remedies prescribed and taken. These will have been studied by scientific methods exhaustively. The next most common chunk will have received some cursory attention, and then you enter into the vast fields of the unknown.
But the fields aren't unknown to just science: they're unknown to cultural practitioners as well. I'm willing to believe that, even without double-blind clinical trials, people were able to find effective treatments with enough trial and error. But "enough trial and error" means "not the long tail". So once you start talking the 1000th most common plant used for medicinal properties, the amount of experience pre-scientific practitioners have accumulated is negligible, and handful of anecdotes at best. Not even enough to ascribe correlation, let alone causation.
So on the head of plants of potential value, I'm betting that they've all been studied exhaustively. On the tail, there are countless plants which could have countless benefits, but essentially no evidence to do anything but pick one at random and hope it cures what ails you. The middle, of enough evidence to suspect or know a plant does something, but not how, is going to be fairly small.
15.7% had been tested in clinical trials; 50.9% had been tested in vitro or in vivo; 20.8% had been subject to a phytochemical study; 12.1% had no studies found in the literature search; 0.5% were "dangerous".
That's not saying only 1/3 of plants had been studied. That is saying that 87% of plants have been studied.
[1] http://onlinelibrary.wiley.com/store/10.1111/j.1365-2710.200...