It's not really a conjecture, exactly. For example, suppose you have created a theory that describes what might happen at very small scales. What happens when you take that theory and "zoom out"? The answer is that you should be able to describe what happens at large scales -- you can "recover" our current understanding of physics from this new theory.
This is how, for example, Einstein knew he'd got his equations for General Relativity correct: he could get Newtonian physics right. Likewise, in an early draft, that conceptually is the same as his actual paper in terms of what's going on, he got the equations "wrong", and knew they must be wrong because he wasn't getting Newtonian gravity out of it.
So it is in this case. We don't know what's going on at Planck scales, and indeed it may even be that nothing special happens there. But we *do* know, or at least we know to expect, that if you get the small-scale right, then it will be an extension of, rather than a replacement for, General Relativity. As a case on point, Loop Quantum Gravity, whatever the hell that is, seems to be running into this precise roadblock right now, ie we aren't getting out of it anything that matches our current understanding.
The point really is that the "...but we can still expect such-and-such" is that, in order to make progress, it's good to establish a sensible starting point. And a very sensible starting point is this idea that our current description, while surely incomplete, is correct as far as it goes. General Relativity is a good description of gravity, that has survived countless experimental checks, that only breaks down in extreme examples -- but in that case, you want to patch it, rather than tear it all up.