> which is totally backwards from a basic scientific perspective
This is not right, because if we have a situation where our theories and observations don't cohere, it's not given whether the theory requires modification or we're missing something in our observations (or both). A classical illustration is the orbit of Uranus being observed in the nineteenth century to be contrary to the predictions of Newtonian theory. Calculations were made assuming the truth of the Newtonian theory and that we were missing something in our observations - the position of Neptune was predicted and it was subsequently discovered.
On the other hand, the orbit of Mercury diverged from the prediction of Newton's theory. Again, a previously unobserved planet closer to the sun was postulated as being responsible, but in this case it really did require a modification to the theory of gravity: general relativity, which accurately predicted the 43 arcseconds per century of perihelion precession by which Mercury's orbit diverges from Newtonian predicitions.
GR has obviously made many other predictions, such as the gravitational bending of light, black holes, and gravitational waves, which have been vindicated.
So there's obviously a problem of the theory and observations not cohering, but whether the solution is a modification of the theory or a new form of matter is not clear in advance, and the latter is not unreasonable and certainly it's not unscientific to make as a hypothesis, to see where it leads.
The difficulty is in coming up with a theoretical framework that retains all the successful predictions of GR while also accounting for the galactic rotation curves.