PharmAlert ranks findings by three things: how serious the risk is, how much the patient's own circumstances change that risk, and how well established the evidence behind it is. A theoretical interaction in a healthy 30-year-old is not the same finding as the same interaction in an 82-year-old with reduced renal function — and it shouldn't look the same on screen.
Overridden alerts stay in the record but move out of the way. The reasoning behind every ranking is visible, because a check nobody can inspect is a check nobody will trust.
PharmAlert supports the decision. It does not make it.