Routing in Wireless Mesh Networks is challenging due to the unreliable characteristics of the wireless medium. Traditional routing paradigms are not able to propose an efficient solution to this problem. Further, Gupta et al. demonstrated that the average throughput capacity per node of a wireless multi-hop network decreases as 1/n, where n is the number of nodes in the network. Recent studies have shown that a cross-layer approach is a promising solution to get closer to the theoretic throughput capacity bound. Cross-layer solutions have been already proposed either for specific TDMA/CDMA networks or for power-efficient routing protocols. These proposals are strongly MAC dependent, or suffer from targeting a steady state offering the best trade-off performance. In this paper, the problem we tackle in a more general context, disregarding the specific MAC and Physical layers technologies, can be formulated as follows: How to design a routing algorithm able to increase the average throughput capacity experienced by Wireless Mesh Networks? Starting from a theoretic result, we analyze the gain that a cross-layer approach can deliver, the metrics suitable to improve throughput capacity, and the power control policy that reduces interference. We take a MAC independent approach, focusing on the general characteristics of wireless links, targeting the improvement of throughput capacity in Wireless Mesh Networks. Our proposal performs path selection and power optimization based on three metrics, namely physical transmission rate, interference, and packet error rate. Performances are thoroughly analyzed and evaluated by extensive simulations, with both TCP and UDP traffic, and compared to other multi-hop routing protocols. For both kind of traffic, the simple heuristic we propose here allows to double the average throughput the network is able to route.