On the asymptotic optimality of alternative minimum-distance estimators in linear latent-variable models