Well, im supprised you need 20secs per image to plate solve. I use 7 secs at bin 3 (lum filter) without a problem, with maybe 2 rejects per 40 images. If you get this failure rate with 20 images, thats 10 good solves, not good, look into that 1st (6 is the min for any result).
Does your map have a good spread across the sky?.
I use
AAG Tpoint mapper BTW (free), ive heard its more user friendly and it shows a picture of your scan map.
The Tpoint polar align data (and adjustment recommend) is always bang on for me.