Process discovery refers to the task of learning a process model on the basis of data about historical executions of process operations, in the form of an event log. Recently, a new family of bottom-up discovery techniques known as eST mining has been established: the eST miner is able to flexibly produce Petri nets from event data, to represent long-term dependencies, and to guarantee a certain output quality. In this paper, we provide formal proofs for meta-properties of selected Petri net classes, and we propose an extension pattern for bottom-up discovery methods able to produce Petri nets of chosen classes more efficiently, by leveraging meta-properties results to prune the search space of model components. We show how this pruning method can aid bottom-up discovery by applying it to the candidate places search of the eST miner.