And yes, baryonic distribution is absolutely a parameter, but it's not a free parameter (or it's a less free parameter) because it's value is constrained to a measurement that is orthogonal to the quantity inferred (light vs rotation curve). Meanwhile, dm density is a free parameter. It could be zero, or, 10x the baryonic mass, or anything in between.
> "DM theory needs to make predictions we can then test"
That is what scientists do: Start from the hypothesis of a dark matter dominated universe, from the beginning (or soon after the big bang), then turn on time and physics (gravity plus gas physics in a computer simulation), and galaxies form as gravity causes matter to clump together. The properties of those theoretical galaxies are testable predictions.
Even so, there are predictions low-parameter DM models seemingly can't make, like: what percentage of galaxies have zero dark matter?
You assert that it is impossible for a model of a dark matter dominated universe to "predict" galaxies without dark matter, but that is exactly what people are looking at here in a large scale cosmological simulation: https://arxiv.org/abs/2202.05836