You can get about the same amount of correctness in Haskell or Rust as you can in typical languages by omitting all unit tests.
This is where people saw a certain sliver of Haskell or Rust community pursuing increased correctness with diminishing marginal return and assume they must represent all of the Haskell or Rust community. But didn't the same communities advertise their languages by that magical feeling of "if it compiles then it probably works"? Plenty of people use only simple language features, write no unit tests, and achieve a similar level of correctness as software in other languages with unit tests. But they aren't as visible because efficient boring software engineering doesn't get blog posts.
Besides that, I actually totally agree with your first paragraph. Not all software needs to be totally correct. But my argument is that if you know you don't need total correctness, it will be less work to achieve a certain level of correctness in Haskell or Rust than in typical languages.