There is a continuum between what is clearly "inside" a network and the rest, such that the interpretation isn't that different as if the result were "inside" the network.
Are they correlated with the behavioral measure or with the network?
The strength of the association of these regions with the RSN is correlated with the behavioural measure. We can't say if these regions are positively or negatively correlated with the rest of the network based on the relationship with the behavioural varialble. For this you can just look at the sign for that region in the group map for that RSN.