A Composite shot with several layers consisting of a Back ground plate, A 3D model, a keyed video layer, light,camera and a point layer of which the 3D model is parented to.
After highlighting the layer one wishes to modify with the mouse. i.e. 3D model parented point layer. Use the mouse to manually adjust the position with the X,Y,Z widget.
If perchance the mouse slightly deviates from it's position on the widget, it will activate and highlight the layer that is below the target layer and that layer will then be moved.
If you do not follow this with a quick CTL-Z , your critical placement if that layer will be compromised.
Why can we not limit mouse induced layer movement to the "Highlighted" layer?