Redmonk looks flawed too. I would expect GitHub to create a major sampling bias in favor of smaller, throwaway projects, and in fact, a quick search(1) on GitHub appears to show 2.2 million projects smaller than a kilobyte that have never gotten a single star. Should those be counted? Do they really reflect anything meaningful about the language's rank?
In my opinion, the only meaningful measure of a programming language worth quantifying from the net are StackOverFlow tags: they can be counted and read and provide a real reflection of the language's community without giving any special preference to language, license, or context.
Everything else is hopelessly subjective: whether or not the language maps well to a particular problem, whether it maps well to the way you like to program, and whether it fits into the ecosystem in which you plan to deploy it. You can't measure those with google searches.
(1) https://github.com/search?utf8=%E2%9C%93&q=size%3A%3C1+stars...