Basic question #4: Robustness of the FDR network inference to larger (but unconnected) graphs

The basic motivation for using the FDR method for inference is to automatically distinguish the significant edges in a given network, in a conservative way.

As a test of the FDR procedure, I will revisit the case of $N=9$, $M=9$. (It now occurred to me that I was not careful about checking for repeated edges in the underlying directed edge construction. Furthermore, it is possible to get $n_i \to n_j$ and $n_j \to n_i$, in which case, the "true" number of inferrable undirected edges will be less than the nominal $M$.)