The problem with the 0.08+ statistics is that they are based on a data set which is fundamentally flawed and biased. 2/3 of the people from the San Diego validation study (where the 0.08 was validated) were above a 0.10 BAC. To put it in easier terms, without even performing the tests an officer could accurately predict someone being over 0.10 with a 66% accuracy rate.
When you look at the three test battery, it is very good at accurately predicting someone is lower than 0.08 when their BAC is say less than 0.04. They are accurate is predicting someone is higher than 0.08 when someone have a measured BAC of greater than 0.10. But when you look at the target zone BACs, where an officer really needs these tests to be accurate (0.06 - 0.08) the accuracy rate of the tests are between 30% - 60% accurate. A subject is 6 times more likely to be found falsely above 0.08, rather than falsely below 0.08. All based on the San Diego validation study data.
And that is from a peer reviewed article "Statistical Evaluation of Standardized Field Sobriety Tests" by Hlastala et all, published in the Journal of Forensic Science May 2005 Vol 50 No 3.