Such arguments are in general hand-wavy and born out of folklore or personal biases and I hate to appeal to the popularity argument, but in the presence of overwhelming popularity of languages that do have type-systems with subtyping, you do need at least some sort of proof that subtyping is bad in one way or another. Even GHC has extensions for subtyping, without being burdened with Java compatibility.
And what are you comparing Scala with exactly?
If you're comparing it with Clojure or other LISPs, well that's like comparing apples to oranges, because there's a world of difference between techniques used in static languages like Haskell/ML and Scala on one hand and LISPs on the other. "flatMap" in particular is not something that you see often in Clojure. You don't quite see many instances of monadic types in languages like Clojure either. Scala's Option[T] or Haskell's Maybe are impractical in Clojure so you'll never see it in Clojure unless you're looking at code that was written by a Haskell developer. Etc...
Compared to Haskell, Scala does have some disadvantages, however even if you're learning it for learning's sake, it's easier to get started in Scala because of the ecosystem ... you've got everything, including capable IDEs like IntelliJ IDEA and people that complain about SBT never tried Cabal, plus with the JVM you can always use Maven or just plain Jars manually copied in a directory.