Let’s say i have a test in a week and i will be tested on 100 cards’ worth of knowledge. Throughout the week I do those cards and then right before the test, I cram-study those 100 cards. As I go through them, the chance of recall for every card I complete begins to drop from 100% at unique slopes (I know the decay of the chance of recall with time is not linear but I wanted to simplify things). The lower the average slope, the lower the average chance of recall for a question on the test, and the lower the test score. Therefore to get as high a score as possible, we want the target retention to be as high as possible to make the average slope as high as possible. However, we still want target retention to be low enough so reviews are still manageable.
Think your retention and your performance as a track & field athlete performance during a season. You want to train to keep on shape (keeping the set retention, 80-90%). During important competitions, the athlete wants to be at peak performance (max retention<100%). So, the coach can plan all workouts and training sessions (Anki review workload) from a period of time before the competition (exam).
Summarizing, you cannot keep a 100% retention long-term, so you focus in having that in the days it matters the most. Using this strategy, you optimize the maximum score chance in the exam vs the manageability of daily reviews.
In Anki specific terminology, you could build a higher retention short-term by filtering decks, decreasing the maximum card interval, etc. Later on, if you still want to remember long-term what you needed for the exam you can change back the settings for that.