To address the issue of influence overlap commonly caused by existing methods in multi-source spreading on complex networks, this paper proposes a redundancy-aware maximal clique centrality method for key node identification in complex networks. Taking maximal cliques as the fundamental computational unit, the proposed method accurately quantifies the higher-order bridging value of nodes as they span low-constraint maximal cliques. Furthermore, Katz centrality is introduced as a global penalty term to identify topologically dispersed and complementary sub-core hubs, thereby achieving dynamical de-redundancy. Parameter analysis experiments reveal the intrinsic correlation between de-redundancy intensity and network structure. Susceptible-Infected-Recovered spreading dynamics simulations conducted on six real-world networks demonstrate that this method effectively overcomes local spreading interference, achieving the optimal standardized steady-state infection scale across various intervention ratios. Numerical results, supported by analyses of the monotonicity index and average shortest path length, confirm that the proposed method exhibits exceptionally high node-ranking resolution and spatial dispersion.