In the non-cooperative case, among the studies of efficiency evaluation of serial systems using data envelopment analysis (DEA) method, the excellent leader-follower or non-cooperative network DEA model can provide a unique evaluation result for the serial system and their internal sub-systems. However, the premise is a given leader-follower order that reveals which stage is more important for improving system's efficiency. Different leader-follower orders yield various evaluation results and not necessarily all systems obtain their best efficiency in the same order. Furthermore, existing studies on network DEA mostly look decision makers completely rational, which is not in line with the bounded rational behavior in practice. This study focuses on the question of "In the non-cooperative case without a given leader-follower order, how to evaluate efficiencies of the serial system and its internal sub-systems". To answer this question, we propose a novel approach for measuring efficiencies of the serial system and its sub-systems based on a lexicographical optimization approach and the prospect theory, considering the bounded rationality of decision makers and all leader-follower orders. This method can provide a unique, comprehensive, and comparable efficiency composition result. The experiment with the data of 14 electric power companies is used to validate this method.