if the solution is faster than random it could still be a real solution on a quantum computer.

well, it's slower than random

“recovers every reported private key at statistically indistinguishable rates from the IBM hardware runs.”

Did that mean success rate from multiple runs or speed for a single run?