Mathematical modeling of complex biological systems has been shown to lead to contradictory results. Two completely different approaches are compared using a highly dynamic system of the adaptive immune response with medical relevance. In germinal centers, high affinity antibodies are newly generated from antigen-activated B cell clones. The encoded antibody is mutated and high affinity clones for a specific antigen are selected to survive, giving rise to affinity maturation of antibodies. Here, the general assumption that competition for antigen held on follicular dendritic cells is responsible for affinity maturation is shown to be unlikely. This finding is based on the investigation of eight different selection mechanisms. A realistic model on selection mechanisms leading to affinity maturation is developed. It is found that negative selection of B cell, a B cell refractory time for binding antigen, and competition for T cell help, have strongest impact on affinity maturation. Antigen consumption is demonstrated to have impact on the kinetics of the germinal center in the late phase of the reaction and is hypothesized to be responsible for the termination of the reaction. These results are consistently found with agent-based and ordinary differential equation approaches and can, thus be considered to be robust. Experimental tests of the predicted selection mechanisms are proposed.