Priority value is an important indicator for Non-Axiomatic Reasoning Systems (NARS) in allocating resources to tasks to process them. This paper identifies that the way existing NARS implementations handle priority does not always make priorities work according to the principle of resource allocation in NARS. Specifically, 1) the processing frequency of high-priority tasks may be lower than that of low-priority ones; 2) the relationship between priority and the frequency of task processing is not approximately linearly increasing. This paper proposes a mathematical model for the problem and provides a concrete solution. Through a series of comparative experiments under various task input conditions, we analyze how different implementations translate priority into processing frequency. The results demonstrate that the method proposed can accurately make priority work according to the conceptual design in most cases. The code is available at: https://github.com/MoonWalker1997/ImprovedBag