“…There are two properties of H(G) that are important for us. The first one is that any assignment of probabilities h[H(G)] ∈ P[H(G)] can be implemented by suitably choosing assignments p(G) ∈ P(G), x(G) ∈ P(G), y(G) ∈ P(G) and z(G) ∈ P(G) for, respectively, {e k } n k=1 , {x k } n k=1 , {y k } n k=1 and {z k } n k=1 , and assignments of probabilities a(K 2 ) ∈ P(K 2 ), b(K 2 ) ∈ P(K 2 ) and c(K 2 ) ∈ P(K 2 ) for, respectively, {a 0 , a 1 }, {b 0 , b 1 } and {c 0 , c 1 } (K 2 is the complete graph on two vertices; the graph of exclusivity of the events of tossing a coin) [10]. Therefore, the largest P(G) allowed by assumptions 3.1 and 3.2 can always be obtained from the largest P[H(G)] allowed by these assumptions by suitably tracing out its elements.…”