With the growing demand for robotic solutions in the industrial separation of recyclable waste, the need for advanced grippers has become increasingly critical. This study presents a custom-designed and manufactured gripper that accommodates three different suction cups. It features a software-controlled mechanism that dynamically selects the suction cup to be used for implementing each pick. Additionally, due to the highly deformed shapes of recyclables, RGB-D perception is utilized to identify the area on the recyclable surface that is most suitable for suction gripping. Then, by analyzing the spatial properties of this area, it is possible to select the cup that is expected to accomplish the most feasible and stable suction grip. A set of comparative experimental evaluations conducted with a Cartesian robot demonstrate the effectiveness of the proposed approach. By combining contact area identification with optimal cup selection, significant improvements in object picking success are achieved, compared to the conventional practice of using fixed-size suction cups.