We investigate a procedure to reduce the contraction cost of tensor networks for quantum circuit simulation by looking at equivalent graph-like representations in the ZX calculus. We use treewidth as an objective function, which relies on its exponential relationship to the contraction cost as proved by Markov and Shi. The initial proposal of this method is due to Cam and Martiel, where it was applied to optimise the simulation of 53 qubit Sycamore circuits. We extend the study by looking at contractions of smaller circuits with arbitrary all-to-all connectivities. We find that the method seemingly functions just as well with arbitrary connectivities, however an improvement in the contraction cost is not guaranteed, particularly on smaller and heavily partitioned circuits. We also find that larger partitions result in bigger gains at the steep cost of additional runtime. Further testing is required to determine characteristics of circuits such that the optimisation is likely to be beneficial.