Elixir made some solid gains this quarter, and it's based on the BEAM VM, so not JVM/CLR, and not great c/c++ interop. It has great concurrency support, and is widely regarded as being rock-solid.
Haskell fell one spot, but it's still in the top 20, and it's not based on the JVM/CLR, and doesn't have great c/c++ support. The library support is similar to the situation in Python 3, and I've written plenty of performance sensitive Haskell code that's within 2x the speed of c.
It's a big world out there, and there are so many great languages. The JVM/CLR definitely provide a nice head-start, but they're not the only way to build a good language.
NIFs work quite well even if htey have quirks and are limited. And with ports and Dirty Schedulers, you have other solutions for more complex and longer things.
Not perfect, but there is a good story here.
I'm not quite sure what you are expecting. It is a tradeoff they explicitly made.
I've yet to see another platform come close to rivalling the JVM ecosystem. It is a powerhouse.
I think it's a great tradeoff. From a team and organisational productivity point of view unless you are based on either the JVM, the CLR, or have fantastic c/c++ interop you are dead in the water.