This claim may be true, but I doubt it’s true simply because of the graph on the left. The data needs to be normalized based on the quantity of code written in each language. What we’re looking for is the concentrations of “wtf” within a comparable amount of code (e.g. lines, files, expected bytes-per-man-hour, etc).
By an order of magnitude!
The WebKit has a library "Webkit Template Framework" (WTF). It is a huge part of the WTF in C++.