Angular unconformities present a major challenge for this methodology. An angular unconformity is a horizon-like feature that has two dips calculated everywhere. One dip for the layers above and another dip for the layers below. These can be handled in different ways.
Perhaps the simplest way to handle unconformities is to break up the cube. If an unconformity is already identified, then the data can be broken up into different cubes, flattened separately, and then recombined.