I tend to agree with tracstarr's logic. My understanding of these ADHRS is that they work using solid state accelerometers sensing acceleration of gravity and motion. If that is the case, then using tracstarr's example, a sensor mounted off the axis of rotation would sence more motion and therefore greater acceleration during that period of time. As soon as the motion stopped (you reach a point where your pitch or roll is now stable and not changing) the sensor would sense a stable pitch or roll angle and would read the same no matter where it was mounted. So if this logic is correct (and I am just throwing it out there, not claiming it to be absolute) there would be some inaccuracy in the pitch or roll, but only while the motion was occuring. A relative slow motion, or where the sensor is mounted not too far from the axis, would probably have a short lived error that is not even noticeable. Maybe that's why the mounting location is not considered an important factor. Bottom line, don't overthink it like I am attempting to do.