That 20% number sounds high considering the information was most likely not definitive. DNA sure does add a wrinkle. Or gnashing of teeth.
This article references many studies. Apparently some of them ranged as high as 30% error in paternity, but in their broader assessments it appears the “normal” range is much lower in the 3-4% range.
http://www.canadiancrc.com/articles/BBC_Paternity_Fraud_Study_UK_John_Moores_University_10AUG05.htm
I have heard the figure of 10% a lot, but 20% is perfectly credible to me.