This is fascinating, but it's nothing to do with whiteboards, and everything to do with individuals making technical assessments of other people. The title of the article is:
You will randomly bomb technical
interviews. Everyone does. Here's
the data.
That's more accurate that the one given here.
It would be a good idea to look into data from interviewer side. In my experience I found some people provide much better interviewing environment and some people make candidate nervous.
I have the same problem with these analyses from interview.io every time I see them: they are making many of their inferences by fitting linear models to data that is non-linear. Is getting a 4 twice as good as getting a 2? If not, don't use a linear model!