This paper presents a new approach for solving the paradoxical Blackman's Association Problem. It utilizes the recently defined new class fusion rule based on fuzzy Tconorm/ T-norm operators together with Dezert-Smarandache theory based, relative variations of generalized pignistic probabilities measure of correct associations, defined from a partial ordering function of hyper-power set. The ability of this approach to solve the problem against the classical Dempster-Shafer's method, proposed in the literature is proven. It is shown that the approach improves the separation power of the decision process for this association problem.