There was more than that. There was the match and the non-match division, and there was the case where they couldn't determine the gender of the Physician. Also, some of the data doesn't look exactly indistinguishable.

Look at Fig.S2 here http://www.pnas.org/content/pnas/suppl/2018/07/31/1800097115...

It looks like male and female doctors result in the same probability curve for female patients, but male doctors result in a higher probability curve for male patients.