As i read it, it doesn't change actual understanding, but rather removes (some of) the confounding variable of test-taking anxiety. Female students actually understood the material just as well as male students, but they didn't believe that they did, and they didn't test like they did, so for all empirical purposes, they didn't.
The writing task was basically a luminosity exercise: making them aware of their own true knowledge and beliefs through introspection, such that they then had the confidence to take risks on questions they otherwise wouldn't have (assuming the tests were scored negatively on incorrect answers to discourage guessing.)
That is a really interesting question. Given the "control" was also given a writing exercise, I think it opens a lot more questions. If we hold it true that the 15 minute writing assignment makes a difference by itself (questionable, obviously), it would bring up interesting questions about male versus female motivation, where males do better when writing about others' motivations while females do better when writing about their own.
The writing task was basically a luminosity exercise: making them aware of their own true knowledge and beliefs through introspection, such that they then had the confidence to take risks on questions they otherwise wouldn't have (assuming the tests were scored negatively on incorrect answers to discourage guessing.)