For reusable launch vehicles, it is common to separate the guidance and flight control problem into an outer loop and an inner loop. The inner loop uses aerodynamic and propulsive controls to achieve a commanded attitude. The attitude commands are generated by an outer loop. This outer loop uses inner loop commands and control variables to achieve desired positions/velocities. This paper describes improvements and evaluation results for combined inner-outer loop adaptive flight control architecture for reusable launch vehicles. This architecture includes the online training of a neural network to correct for force and moment model errors. The result is a system that can respond to large model errors (significant failure scenarios) in both inner loop attitude control as well as in trajectory tracking. The resulting system can be used to improve performance and/or to increase guidance, navigation, and control system tolerance to model error, failures, and the environment. Simulation results are presented for failure scenarios.