Am I correct in interpreting this as that among 15-year-olds, girls and boys are about equally skilled at doing math, but girls are even more skilled at doing reading than boys are, so it's relatively inefficient for girls to do math-oriented tasks instead of reading-oriented tasks.

With the relative inefficiency might come relative unprofitability, and if the girls perceive and/or anticipate that doing reading-oriented tasks would be more profitable for them, they choose to participate in the market for reading-oriented labor (self-select into that market).