The only physical mechanism I can think of for quasars to have such a large difference in two bands would be at very high redshift ($z\gtrsim6$) where you have the Gunn-Peterson trough completely erasing everything blueward of 912 Å (restframe). But with typical magnitudes of $\sim20$ or so at these redshifts, a color of $20$ that would then mean that the magnitude in the bluest band would be of the order $\sim 40$, which no instrument is able to detect.
So effectively, there would be no detection on the bluest band, in which case the color would be infinite, or rather undefined.
In other words, I think those values must be errors.