You are correct on both points. You can use just the urban coefficients for all urban areas, and for Java, use the second approach of estimating Java-specific quintiles. You should point out that these quintiles are not comparable to either the national nor the urban quintiles.