This is the expected behavior, the overlayDepth allows you to modulate the fall off for signal that is deep below the surface of the object. Since it is hard for a viewer to gauge depth from a 2D image, you have some control over how depth influences opacity. For the figures you are showing with explicit cutouts you may also want to explore the "OverlaySurface" - it also allows you to control for the depth of the signal, but generates a painted surface appearance.

This reflects the tradeoff between precision and interpretation:
  https://onlinelibrary.wiley.com/doi/full/10.1111/ejn.14430
In general, I think the combination of 2D slices and 3D renderings helps individuals understand the spatial relationships of observed signals.