"Well typed programs don't go wrong" only for sufficiently narrow definitions of "go wrong". They don't go wrong by doing undefined behavior. They can still go horribly wrong by doing unintended behavior, though. Type safety doesn't eliminate all bugs, just a subset of them.
Depending on your type system, though, the subset that is eliminated can be larger or smaller. In Java, it's a smaller subset; in Haskell, it's a larger subset (because e.g. the type system statically rules out the possibility of null), and in a dependently typed language like Coq or Agda, it's a ridiculously large subset of bugs. It depends greatly on your particular type system.
I'll agree that that point may not have been highlighted quite as strongly as it could have been, but that idea is fully included in the original post, and you can view the purpose of examining so many language's type systems as comparing & contrasting the specific "go wrongs" that their type system addresses.
