Why ignore current event information when predicting anomaly？

[cycl2018]

Thank you for providing an interesting anomaly detection method. However, I noticed that when the model predicts the node anomaly state, it finds the historical neighbors but does not include the current event. This seems to ignore the current event information. However, based on my understanding of this task, the label of the node represents whether the state has changed after the current event occurs, which means that the current event should have very important information. Is this just a mistake or are there other reasons for this?