(I work for Mozilla) It captures the page cropped to the bounding box of the element, so any background content would also be captured. The results would not contain transparency.
You would see the element under it assuming it's possible. What this does is finds the bounding box cords and takes a full page screenshot and crops it to that region.