There was a model based on the concept of Transition Probability Matrix (TPM). The TPM was built on obligor ratings on quarterly basis.
The deal breaker was that some obligors were given a PD rating from different models. More explanation: Suppose an obligor A was rated by model B in quarter 2007Q1 then it was observed that in 2007Q2 it was rated by model C.
There were many such incidents.
I will keep on updating this page.