I´ve got a problem to quantify a rider peak (100% API peak, 0.1% impurity, poorly resolved, sitting on the tailing side). Who has any idea how to minimize the error for quantification of the small peak ? Vertically dropping leads to a too high result, tangiental integration to a too small result (peak area).

Try working on the method to fully resolve the tailer and then develop your integration method to do what you need.