There's no contradiction. It makes approximately the right predictions, but it's completely wrong as a statement of how things actually are.
A very clear example of such a hypothesis would be one concerning identity. E.g., who was it who committed the murder? The hypothesis that Joe Smith did it might make lots and lots of correct predictions and yet still be completely wrong (because it was actually someone else).
> It makes approximately the right predictions, but it's completely wrong as a statement of how things actually are.
You're assuming that we know "how things actually are". If we know that, how do we know it? If your answer is "because we have a theory that makes correct predictions", how do you know our best current theory gets everything correct? It has to be literally "everything" in order to justify your statement that we know "how things actually are"; anything less than perfection from our current theories does not justify your claim.
It should be obvious that we do not know that our current theories are perfectly correct. All we really know is that they get more predictions correct than our previous theories. But that still leaves room for our best current theories to make incorrect predictions--we just don't know which ones they are (yet). And if they make any incorrect predictions, then there could be some other theory that makes more correct predictions, but models "how things actually are" very differently from our current theories. After all, that has already happened, when we came up with our best current theories. If it happened once, it can happen again. And if it happens again, then all your claims about "how things actually are" are, by your own definition, wrong.
>You're assuming that we know "how things actually are".
No, I'm only assuming that we sometimes know how things actually aren't. In other words, we can sometimes be pretty sure that a particular theory is wrong even if we have no confidence that our current best theory is correct.
A very clear example of such a hypothesis would be one concerning identity. E.g., who was it who committed the murder? The hypothesis that Joe Smith did it might make lots and lots of correct predictions and yet still be completely wrong (because it was actually someone else).