The differences in callback rates were not statistically different between the women and the lower class men. The only statistically significant different was higher class men and everyone else.
I didn't do the stats on it, but just looking at the graph in the article 5 times as many lower class women received callbacks than lower class men. What was the p-value on that?
Instead of "5 times as many" look as "a difference of 4 samples" - in an environment where the expected sampling noise is something like +/- 4.
"Statistically insignificant" means just that - this data is not sufficient to conclusively say if lower class women receive less callbacks than lower class men; maybe they do, maybe they don't.