It's the classic failing of everyone's casual "alternative to dark matter". It's not X because we can't detect it, what if instead it's an undetectable form of Y?
Far too many people get hung up on the idea that in science what things "are" matters: because it doesn't. What matters is predictive power for future observations - if whatever you're proposing doesn't predict anything different to "invisible matter" then it's exactly as bad an explanation as "invisible matter" which so far is a perfectly good explanation for current observations.
Then you just get fancy regression formulas. With enough layers and "bars", epicycles "work" in terms of predictive power. It's usually better to find out the underlying mechanism rather than merely satisfy prediction. Otherwise we'd still be using epicycles. Plus, gravity theory has Occam's Razor over epicycles.
Right but that's the point: there's nothing wrong with "invisible matter" as an explanation if it explains everything. Even if it's actually not invisible matter, provided it behaves like it at all times, then that's good enough until it doesn't.
The universe has no requirement to satisfy the day-to-day intuitions of African-plains originated upright apes, with senses optimized to spot large predators, in how it works. Quantum mechanics is a great example of that.
Far too many people get hung up on the idea that in science what things "are" matters: because it doesn't. What matters is predictive power for future observations - if whatever you're proposing doesn't predict anything different to "invisible matter" then it's exactly as bad an explanation as "invisible matter" which so far is a perfectly good explanation for current observations.