It's possible that your declination and/or elevation is off, but you can get it to track either east or west (but not both) by compensating errors. For example this can happen if at the zero position the dish is not pointing exactly true south. If the dish mount on the DG-380 isn't centered, this can also cause a similar problem. Getting the dish mount exactly centered on the DG-380 shaft is the first order of business. I am assuming everything is plumb, because you've checked that.
Next you set your true south satellite for the best elevation and then choose an extreme satellite to the east, say. Drive the dish with USALS to the correct position, but don't mess with elevation. Rotate the mount on the pole for the best signal and mark the mount position on the pole. Now do the same on the west side, hopefully by roughly the same orbital offset from your true south sat. Readjust the mount so it is exactly in between the two marks. Readjust the elevation on the true south sat and adjust the declination to get the extreme sats. This should get you in the ballpark.
I have the same motor and dish as you and I'm not that excited about the precision. The DG-380 mount to the pole wasn't strong enough to get it precisely plumb without freezing the u-bolts. I got some grade 8 u-bolts that were a lot bigger and that taught it a lesson. I also experienced some slippage of the dish mount on the motor shaft during high winds. I shimmed it with some split 2" PVC schedule 40 and used heavier u-bolts for that, too. That pretty much cured the sloppy mechanical stuff.
I also discovered there was a slight bias error in my DG-380 zero position. Uncompensated this could cause the trouble you're having in tracking the arc. It wasn't immediately obvious the best way to correct this mechanically. Fortunately for me I'm using some Linux software I wrote to drive the motor, so I just applied the offset in my software before calculating the positioning command. I later bought another DG-380 as a spare and that unit's zero was pretty much dead-on.
Finally you should be aware that if you use USALS to calculate your motor angles, the algorithm apparently has increasing errors as you move farther off true south. My analysis indicated they did not take declination into account in their calculations, which is pretty sad. I wrote a thread about this a number of months ago. If you're tracking the arc but USALS puts you in the wrong position by an increasing amount off true south, this could cause problems, too. But not to the extent of what you are seeing at the moment.
.
