One of the complications described by the author is performance. Personally, I find stdlib graph libraries extremely useful even if their performance is poor because it’s often the case that my dataset is small enough, and even if performance turns out to be an issue, first spending time on the problem with a underperforming graph library is a very worthwhile exercise before trying to write my own optimized implementation. By analogy, many programming languages are far from being the fastest, but they can nevertheless be very useful.

That said, I’m not surprised performance came up in interviews with experts; they probably have tons of interesting performance-related stories to tell from their extensive work on graphs.