Every few years, the International Planning Competition (IPC) challenges Artificial Intelligence (AI) planning systems with various problems to advance research in this field. These systems utilize knowledge models, also referred to as benchmarks in the context of the IPC, to acquire the necessary knowledge for generating solutions to specific planning problems. Due to the wide use of such systems in the real world, realism should not be neglected while developing knowledge models; otherwise, produced plans might not be applicable or subpar. Therefore, by modeling the benchmarks realistically, the results of the IPC could become more meaningful for real-world applications. To qualitatively assess realism, we apply a conceptual framework known as the realistic-aspects framework to the benchmarks used in the Hierarchical Task Network (HTN) track. This track focuses on the HTN planning approach, which is commonly used in practical applications and decomposes complex tasks into more manageable subtasks using a hierarchical structure, enabling efficient problem-solving. We further enhance this framework by modifying specific components. Additionally, we develop metrics based on this framework to measure realism quantitatively. The presented evidence suggests that many knowledge models, in general, and those used at the IPC in particular, fail to reflect certain aspects present in real-world domains. The improved version of the framework, along with the synthesized metrics, reinforces those findings. The ability to measure realism, both qualitatively and quantitatively, offers planning engineers a practical way to develop realistic knowledge models. Consequently, this could improve the practical relevance of IPC results and encourage broader adoption of AI planning systems in various real-world domains.