Hi Peter,
In terms of how LS-DYNA considers energy balance after element erosion, I will refer you to the link below where more information is available.

However, for your specific case, it appears to me that due to localized deformation (via shear banding, for example), a layer of elements in the projectile satisfies the condition for element erosion while the surrounding material is still below the threshold for erosion. As a result, a part of the projectile (shown with red circle) is detached and travels through the target. I think by setting enmass to 1, this issue could be resolved and it hopefully be better in line with what happens in reality.