The survey methodology is biased toward old technologies, from the era where you sent HTTP headers that advertised the stack you were using. Modern stacks often hide that info. (It's also biased toward techs used in blogs, since when it detects a tech used on blog.example.com, that tech is counted for example.com)
Also it biases towards smaller organizations which don’t follow security guidelines saying that you should disable those banners. There’s a real breadth versus depth discrepancy here.
