Here's a set of histograms of lambda error for all the points in my VE table as extracted from a recent data log.
(error is defined as actual lambda - target lambda)
Note the errors have zero mean, and a fairly small variance, so lambda correction works very well, and my VE table can't be that far out either...