Recently, many multi-channel protocols and architectures have been proposed to exploit the non-overlapping channels, especially in MANETs and Mesh networks. With the development of modulation and encoding technology, the diversity of data transmitting rate could lead an increment of discrepancy between channels. The contribution of this paper is threefold. First, when allocating channels with quality discrepancy, whether the competition on the high-quality channel should be avoided depends on the degree of the discrepancy. A channel competition model is built to quantify the discrepancy. Second, by modeling the channel allocation problem as a game, we found that when the relatively poor-quality channels are good enough to share the other channels burden, a mixed Nash Equilibrium is needed to improve the fairness of allocation mechanism. Third, a mixed strategy-based channel allocation mechanism (MSCA) is proposed to improve the overall performance by introducing randomcity into channel allocation. Simulation results on NS2 show that, by allowing the users to adaptively changing their strategy, MSCA can improve the throughput by 10%-15% and reduce the end-to-end delay by 20%-30%.