“Martin Lercher estimates that 95% to 99% of outlier data points in scientific experiments are the result of simple mistakes.”