It seems kind of wrong to be giving a reliability score without any sort of testing, and labeling it as such. It should probably be a ‘predicted reliability’ rating instead.