Didn't Hosokawa prove (1958, Osaka J. Math) that the Alexander polynomial of a knot can be any integral Laurent polynomial p(t) such that p(t^-1) = p(t) and p(1) = \pm 1? If that's right, then according to Hosokawa, 2t^{-2}+t^{-1}-7+t+2t^2 would be the Alexander polynomial of a knot, contradicting this conjecture. It's been a while but I think you construct these knots very explicitly using ribbon diagrams -- Rolfsen's knots and links, also Kawauchi's big survey book should have the construction.