I don't really know the answer, but I suppose I would first start by trying to *disprove* the conjecture.  After all, it has only been verified for graphs up to order 5.  The obvious counterexamples I would check are large cliques and large anti-cliques.  

So, do there exist arbitrary long sequences of demi-primes that are pairwise co-prime?  What about arbitrary long sequences of demi-primes such that each pair has a common factor?  

The number theorists can feel free to chime in here anytime.

If those don't work, then some other candidates for counterexamples would be large matchings or large cliques with an isolated vertex.