This paper focuses on the development of a new revised Desired Compensation Adaptive Law (DCAL). DCAL is a model-based adaptive control strategy consisting of three main parts: (i) an adaptive feedforward term, (ii) a linear PD feedback term, and (iii) a nonlinear compensation term. In order to deal with highly nonlinear dynamic systems characterized by their abundant uncertainties and parameters variations, we propose to revise the original DCAL control law by adopting adaptive feedback gains depending on the system state errors. Besides, DCAL controller is known for its robustness against measurement noise thanks to its desired compensation design, but a large amount of external disturbances are still not compensated by such a design. Therefore, the proposed DCAL with adaptive gains (DCAL-AG) is extended with a sliding-based term to further improve its robustness and the overall performance. A model-based robust adaptive feedback controller appropriate to the control of nonlinear systems in real-time applications is thereby obtained. To demonstrate the improvements brought by the proposed control strategy, numerical simulations have been conducted on a Delta-link parallel robot named T3KR in a "Pick-and-Throw" application task at different operating conditions.