I am really sorry that it has taken almost a year and a half for me to respond to this and for the poor documentation of 3dECM.
If you are using the full power method, described in Lohman et al. 2010, then the first subbrik corresponds to ECM calculated on a binarized graph and the second corresponds to ECM calculated on a weighted graph. The binarized graph is created by setting the weight of all connections with correlation above the specified threshold to 1, and all those below to zero. If a threshold is not specified on the command line, then it defaults to -1.2, in which case the binarized graph is 1 for all possible connections in the graph.
If you are using the fast method, the first subbrik corresponds to ECM calculated on the weighted graph and the second subbrik should all be zero.
I do not remember why we did things this way and now can see a variety of problems with it. If I can find the time, I promise i will straighten everything out.
Cheers,
Cameron