Well, Moffat has developed four or five different theories of gravity himself. NGT is just the first of these. The most recent, Scalar-Tensor-Vector Gravity (known as MOG - MOdified Gravity), like TeVeS, is a tensor-vector-scalar theory, and attempts to explain the flat rotation curves of galaxies without calling on dark matter. As far as I know it’s viable and also predicts the accelerating expansion of the universe without dark energy. Well if the only reason to propose dark matter was to explain the observed galaxy rotations, and the dark matter particle could not be detected, then we might prefer an alternative to GR, like MOG for example. But it’s not as simple as that. The presence of dark matter is inferred for more than one reason (eg, gravitational lensing of galaxies, the observation of the Bullet Cluster, the need for a critical density of mass-energy to make the universe geometry flat, and so on). Next, it is far from being the case that every potential dark matter particle has been ruled out - there are still many potential particles to attempt to detect. Then, MOG is considerably more complex and less elegant than GR, adding a massive vector field and three scalar fields to the usual relativistic tensor field. It feels a little like adding epicycles - add enough free parameters and you can model anything. But you’re right that ultimately, if dark matter can’t be detected directly, we need to seriously question whether GR is a good model for gravity at galactic scales.