Current methods for causal discovery typically report a single directed acyclic graph (DAG). Through an example, I hope to convince you that this might not be the best practice. In fact, depending on how two DAGs intersect and the local geometry at the intersection, the hardness of this problem can vary dramatically.