Annabella, you're absolutely right that there is a similar problem in health care, where governments under-spend on genuine public goods or goods with externalities! such as immunization, and therefore over-spend on private goods such as clinical and curative care. I'm not sure World Bank conditionality will solve the problem (or any other problem for that matter), but it's certainly worth highlighting the extent to which government spending patterns are biased in favor of the non-poor.
On your second point about the externality associated with girls' education, to the extent that the main beneficiaries are the young woman's children, the beneficiaries are within the same household, although not the same individual. It's also interesting to note that this effect does not seem to depend on the quality of education but simply on the girls' being enrolled in school. There is some work on Bangladesh that shows that a girl's having completed secondary school has a significant, negative effect on child mortality, even if the quality of that education was poor. This only strengthens the point that the learning benefit of education is largely a private good.