There's a subtle distinction between these two statements. Any theory that diverges from our currently accepted theories (like general relativity) will only st

There's a subtle distinction between these two statements. Any theory that diverges from our currently accepted theories (like general relativity) will only start differing at sufficiently high energy. There's usually a dimensional analysis type of argument that will tell you roughly when that will happen. For gravity that's the Planck scale. Now, this is true for any theory of quantum gravity, including loop quantum gravity, string theory, or any other theory that you'd consider. There is just no way to distinguish the modified theory from standard GR below the Planck scale. Now, this is the sense in which people say that string theory is untestable. But again, so is any other theory of QG. So, if your argument for not studying string theory is that it doesn't make testable predictions, you are actually saying that you should not study quantum gravity at all. Interestingly, though, string theory (unlike things like LQG) is actually a theory of more than gravity. So, unlike these other versions, string theory at least has a hope of being testable. The problem that we're currently facing within string theory is that while there's essentially a unique theory at high energy, at low energies like the ones at the LHC, the theory has multiple solutions (sort of like a quadratic has two solutions). The issue is that the number of these solutions to string theory is too large to go through and check one by one (people throw around the number 10^500). I think this is why people sometimes say that string theory is more of a framework than a fundamental theory. It's possible that someday wed figure out which the correct solution is, who knows. At the same time, constructing a quantum mechanically consistent theory of gravity is very very hard and, in fact, the only example we have of such a theory is string theory. That's the sense in which people say that string theory is a successful theory of QR but still isn't able to make any definite predictions.