The job shop scheduling problem with makespan criterion is valuable from both practical and theoretical points of view. This problem has been attacked by most of the well-known meta-heuristic algorithms. Among them, tabu search has emerged as the most effective approach. The proposed algorithm takes advantages of both N1 and N6 neighborhoods. N1 neighborhood is used as a path relinking procedure while N6 neighborhood with its guideposts is applied in a tabu search framework. In addition, a method is presented for updating the topological order, heads and tails in N6 neighborhood. The algorithm is tested on standard benchmark sets, outperformed all previous approaches (include i-TSAB) and found six new upper bounds among the unsolved problems. Furthermore, we have tried to collect the newest upper bounds for the other problems.