Air traffic controllers (ATCO) work in an uncertain environment where they operate traffic with a double objective of minimizing flight times while ensuring safe separations, i.e., detecting and solving trajectory conflicts. They constantly face uncertainties because of variable weather conditions, aircraft speeds, or pilots response time. The development and acceptance of decision support tools to help controllers perform a safe separation must account for these uncertainties and align with operational practices to ensure satisfactory user adoption. In this paper, we build upon a previously published dataset of actual conflict resolutions based on historical ADS-B data and flight plans. In this previous work, we proposed a heuristic to detect deconfliction situations among aircraft persistently deviating from their intended route. Here, we extend our approach to previously overlooked areas of the data. In particular, we apply a KNN-Median regression approach to additional explanatory variables, and gain more insight in the way ATCO cope with potentially conflicting traffic. The most significant improvement is the ability to extract deconfliction situations leveraging "direct-to" instructions.