Haptic shared control is used for various remote operations in unstructured environments, but few studies have focused on human-centered experiments. Because the control effort, comfort, and fatigue experienced by human operators are significant factors in evaluating the performance of shared teleoperation systems, human-centered methods should be considered in experimental design. This study evaluates and analyzes the performance of a shared teleoperation system developed by the authors in a previous study for performing maintenance and repair tasks in nuclear power plants using human-centered experimental methods. This system applied three elementary technologies, namely egocentric teleoperation, virtual fixture, and vibration suppression control. To evaluate the developed system, we designed six experimental cases that combined elementary technologies. Twelve participants performed four repeated experiments in each case. Moreover, the teleoperation task was a maintenance and repair task in an environment imitating a nuclear power plant. Seven performance metrics were defined from the perspectives of the slave, master, and human operators. The evaluation indicators were defined as the completion time, trajectory length, and interaction force in terms of the slave; pressed button, moved distance, and steering angle in terms of the master; and NASA task load index and overall workload in terms of the human operator. As indicated by the experimental results, the developed system performed better than the conventional teleoperation system. In this study, we focus on human-centered experimental methods and evaluation for a developed teleoperation system that requires significant control efforts from the human operator.