I reevaluated the correlations of Friday. (They were in error since the mean values of the parameters had not entered the MAPLE equations). The new version http://www.slac.stanford.edu/~elsen/amb_new.pdf investigates the case of zero and 0.8 correlation between lambda_1 and Lambda_bar. Consequently the assumed uncertainty of 0.09 (value of draft) translates into a 0.09 GeV (0.15 GeV) uncertainty on m_b for correlation=0 (or 0.8). The actual contour shape in the parameter a varies with the assumed correlation. However, since the resulting uncertainty from a is smaller it is probably not worth the effort to go into detailed simulations. The proposal is hence to increase the theoretical of m_b to 0.15 GeV, such that it is by far the dominant error with ~20%... Eckhard