First, a methodological point: the very high modularity that appears in the first picture (which represents all co-occurrences) is an artifact of SSNA itself. The cliques like the one highlighted in blue are induced by a single, dense post which is coded with 5-10 codes, which then, by definition, form a clique. On the other hand, the fact that the graph is connected (all nodes in the same in connected component) is not artifact of SSNA. The graph that has a more meaningful interpretation is the second one, where only the edges between codes that co-occur at least twice are preserved.
It’s just an intuition, but I expect this finding to be representative of the broader debate, not just of the NGI Forward community.