Bullet cluster came much later and indeed the mass/visible matter discrepancy was discovered specifically because they were looking for evidence for dark matter. It’s just one piece of additional evidence for the theory that dark matter is specifically matter.
Seen this way it seems DM theories are a failure of applying Occam's Razor on a scientifically complex topic. Whereby our current tools and theories cannot explain some observed discrepancies, we aim for a more complicated explanation for them (special types of matter) rather than limitations of our current techniques.
At the same time science doesn't progress by just shrugging and saying "we can't know", so maybe it's an inherent conflict of interest at play here?
More complicated? That makes zero sense and is some sort of a strange cognitive bias against particles (or possibly against things we can’t directly see?)
When it was found that experiments didn’t support Thomson’s "raisin pudding" model of the atom, the solution was not to modify electromagnetism but to introduce a hitherto unknown particle, the proton. When it was found that the mass of a nucleus wasn’t simply the sum of the masses of the protons, the solution was not to modify theories of inertial mass and gravity, but to introduce a hitherto unknown particle, the neutron. When it was found that in a beta decay some of the mass-energy was unaccounted for, the solution was not a modification to Einstein’s equations but to introduce a hitherto unknown particle, the neutrino, whose existence was only experimentally proven quite a bit later.
All our evidence from distinct sub-disciplines of astronomy suggest that some specific well-constrained type of matter that interacts via gravity but not via EM is needed to make predictions match observations. No competing hypothesis has been proposed that is as good a match. Scientific method 101 says that if something walks like a duck, swims like a duck, and quacks like a duck, we should assume that it is indeed a duck until and unless additional evidence rules out the duck model.
