State Matching and Multiple References in Adaptive Active Automata Learning