It's not able to do full quantum annealing; it's only able to use stoquastic Hamiltonians on a fixed topology. Even if truly quantum effects were present (which

It's not able to do full quantum annealing; it's only able to use stoquastic Hamiltonians on a fixed topology. Even if truly quantum effects were present (which is unclear at the moment, see Smolin and Smith arxiv posts), it's not known to be a useful quantum computing model. So it's not just an issue of "optimizing output".