Modelling the impact of maternal HIV on uninfected children: correcting current estimates