My point is, the pool is broken down into subelements. These subelements are broken down into groups. In total, the number of groups in each pool equals the number of questions on the test.
This would suggest that, in order to fully test one's range of knowledge, one question is picked from each group.