Process Mining is a novel technology for discovering process-related information from business data, aiming to discover, perform compliance checks, and improve business processes. The discovery of business processes is the first step in process mining. Due to the inability of traditional algorithms to identify task instances with the same name but different execution semantics, i.e., duplicate tasks, this paper proposes a new process discovery method. Its main principle is to leverage the feature of transition systems that can track the pre- and post-execution states of tasks. This feature is used to distinguish tasks with the same name under different execution states. Subsequently, it constructs a set of directed flow relations that describe the predecessor and successor relationships between tasks. Then, it employs an inductive discovery algorithm, Inductive Miner, to transform the set of directed flow relations into a process tree for identifying complex relationships between concurrent, choice, and other tasks, which are then transformed into a Petri net. Experimental results demonstrate that this method not only identifies duplicate tasks but also reduces the number of implicit transitions in the model. This significantly improves the accuracy of the discovered process model compared to the Inductive Miner.