For modeling individual galaxies, that's more or less correct. (Of course, the same applied to things like MOND: "hey, if Newtonian acceleration is tweaked with

For modeling individual galaxies, that's more or less correct. (Of course, the same applied to things like MOND: "hey, if Newtonian acceleration is tweaked with this interpolation function and that free parameter, it will explain the observation".) In a large-scale, statistical sense, the answer is: initial quantum fluctuations (as seen in the Cosmic Background Radiation) + gravitational collapse and fragmentation in an expanding universe. Simulations done under these assumptions have done an increasingly good job of describing the general distribution of dark matter, and the typical statistical distribution within galaxies. This is why it's good to do independent fits of dark matter distributions to galaxies: it provides something you test the simulations against. After all, if the individual-galaxy fits say dark matter generally has distribution X, but the simulation say that gravitational collapse and mixing should almost always produce something different, then you know you've got a problem. "If it interacts with visible matter gravitationally then why doesn't it take on the same distribution?" Because visible matter interacts with itself via radiation and hydrodynamics (e.g., gas pressure). So it gets pushed around in ways the dark matter can't. These forces can be much stronger than gravity in some circumstances (gravity, after all, is the weakest fundamental force), so the gas atoms, ions, electrons, etc. experience forces the dark-matter particles do not.