I don't think I agree. If a model is known to be largely false but in certain situations it provides good enough results and is easy to work with, that model will continue to get used until a better one replaces it. It's the classic problem undergrads have with Physics I vs. Physics II where the second declares all of Physics I false. Really, it's that Physics II is more accurate under more conditions, but that doesn't make what's learned in Physics I useless.

There is a small step from physics I to physics II, accuracy wise. There is a huge step between physics I and astrology, or basic anatomy and energy meridians. To the point of one of those method from each pair not being useful even in limited specific circumstances.

I don't disagree with your points mind you, I'm making a distinction between inaccurate methods that work in their own limited domain with limited accuracy (that is all we have, I agree) and not even wrong methods that sometimes work by accident but don't generalize even within their own stated limits.