I think this is due to the self-reported nature of Glassdoor, and it exposes how this is a flawed methodology. The original point of salary disparity still stands, but any single datapoint here is likely not reliable.