It isn't predictive capacity just in the sense that it will describe how known things happen - it should describe what will happen in a previously unknown situation, which is where experimentation comes in, whether it is contrived or found in nature. Take the theory that angels pushed planets around and that the movement of the stars was governed by the whim of the gods - when a theory came along (Newton's gravitation) that both described current phenomena, and also was able to predict something previously unexpected (the return of Halley's comet) it was a resounding vindication of the theory.
Yes, and the converse is also crucial: For example the Michelson-Moreley experiment observed a phenomenon (or, rather, lack of one) which defied explanation under Newtonian Mechanics. Because Newton's theory is a good explanation there was no way to make minor adjustments to it which could explain the null result. Instead, we got special and then general relativity, which completely changed the explanation to one in which gravitational forces don't really even exist.
To put it another way, what you said is that good explanations have "reach"; they explain more than the phenomenon they were created to explain. Further, they also tell us what those other phenomena are, because the explanation itself implies that reach (though sometimes we don't see all of the implications). And, finally, they are not easily modifiable to account for new observations which don't fit the theory.
This makes explanatory theories far more than simple predictive tools, and is the reason that the empiricist view of science as merely a process for deriving predictive rules is incorrect.