User blog comment:Hyp cos/SCG(n) and some related/@comment-10429372-20140812170828/@comment-1605058-20140812171959

But you have to take into consideration that whole this page doesn't give us "real" lower bound - in great majority of cases resulting functions are much slower growing that corresponding FGH/HH functions. For example, if we stopped at third graph, with e_e_0, we would get lower bound of f_e_{w^w}(something) (I think), which seems completely wrong when looking at next graph.

Of course, if we use actual FS's given by these graphs (that is, a[n]=max{b<a:graph for b has at most n vertices}) then these bounds are perfectly fine.

Also note that these comparisons don't give upper bound, as we can choose graphs out if order.