Update the implementation and improve the documentation. Currently the estimated sensitivity might be too small for the `experimental` score. Add a warning or raise an exception