Predator assumes that when only one CL is in the regression range, it must be the culprit and assigns it a confidence of 50. This assumption is incorrect for uma-sampling-profiler where the single change in the regression range is much more likely than not to be unrelated to the performance change. The uma-sampling-profiler backend should not use this heuristic and should instead rely solely on the other heuristics.