I am trying to de-cluter a set of points features and am not really getting the results I want.

In the image below the points are the the input features and the pink areas shows the buffer to what was output after transformation. The upper left corner looks ok, but in the lower right the overall "shape" of the cluster has changed to much to my liking.

Is there a way to prioritize the removal of features that are deep within the cluster versus those that are along the "borders"?

I included a screen grab of the custom tranformer i created for this, it removes the features with the most overlaps step by step until everything is solitary.

Any ideas on how to preserve more of the cluster shape while removing all "overlaps"?