In addition, some of the repos that have OCaml code may not be recorded as such. Repos where the 'brains' is in an expressive language might be overshadowed by boilerplate from elsewhere.
I think the adoption problem for OCaml is compounded because it suffers from lack of stackoverflow hits for any given errors that you might encounter or any given queries you might have. Searching for something as mundane as "how to read large files in OCaml" leads to just a single hit (Streams at OCaml.org) .
Also, OCaml needs a "recipe/patterns" book-- on how-to get some of the things done the right way in OCaml.
 BTW, a big fan of your work.
However if you find OCaml in the tiny graph on the bottom you'll see that it's steadily increasing at least. Up about 50% in active repos from 2013 to 2014.
Err... based on what exactly? I've been working in the OCaml community for several years now and it's going from strength to strength. Do we have different definitions of 'strength' or something?
Also, another thing that's peculiar with OCaml is that a lot of libraries are LGPL 3+ which makes it that much harder for corporates to adopt. And sometimes, alternatives to certain libraries are either hard to find, or are not actively maintained. It could also be that, I have been looking in all the wrong places.