I'm just not able to cook up an objection to picking 3 roots in the context and compiling the union of those roots subgraphs