Without statistical moderation, my understanding of the system is like this:
Say student A was ranked first, B was ranked second, C was ranked third. If student A got 70% for the exam, B got 80% for the exam, and C got 90% for the exam, of course they would maintain those as their exam marks (these do NOT change). However, for SAC marks, and I repeat, without statistical moderation, since student A was ranked first, student A would get the first ranked EXAM score as their SAC score; so, 90%. This keeps working so B inherits back his/her own exam score as a SAC score as he/she was ranked second in both the SAC and exam rankings, and C would take A's exam score as his/her SAC score since he/she was ranked third for SACs, and A was ranked third for the exam. Statistical moderation will affect the exact raw SAC scores obtained in the end, but this is a rough guide. Hence, as long as you maintain consistency on the exam, you won't be affected by a weak cohort at all. Someone correct me if I'm wrong here of course...this is just what I vaguely remember from an initially broken understanding).