So the answer is now Crystal Clear to me after seeing results of DC tube response on the following link:
http://www.buildlog.net/forum/viewtopic.php?f=6&t=864&p=6549#p6543It wouldn't matter if your belt had zero slop and your machine is built like a tank. If a typical DC tube can have an on-off delay of up to 2ms, then at 30ips this can result in a position error on the engraving of up to .06" between the sweep right and the sweep left. THAT is why X-unipolar gives a better result than X-Sweep.
So compensation can be done in software if the controller allows, but since the delayed response depends on tube age, temperature, power supply, etc., This value could change over time. I think I'll stick with X-unipolar engraving since the delay errors are always in the same direction and therefore invisible.
30ips=30mil per ms.
I apologize for bringing it up here, because "This" phenomon is not the cause of the problem that was stated at the start of the thread. I agree that there must be a lost step or something that causes the non-circle home position error.