Nah, as others have pointed out, what you do is run the Shor's algorithm, then verify it. If it's wrong, run Shor's again. If it's right, you know you have the factorization. In this way, you can be 100% sure that you've correctly solved the problem, even if Shor's only provides the correct answer some percentage of the time.
What I don't fully understand is why 48% makes this impractical. Having not read TFA, the only way I can imagine that would be the case is if somehow not having exactly a 50% chance of getting the correct answer means that the algorithm doesn't scale correctly. Even only being correct 10% of the time would mean that you could break RSA much faster than you can without quantum computers. I suspect that was some bad editorializing.
What wouldn't be practical under these conditions is factoring larger numbers. You need more qubits for that. Nevertheless, this is a nice stepping stone towards high-qubit computing.