Would you get better results from series data if you either did a weighted average by noise level of the data (similar to EFA, which produces a concentration weighted profile) or first did a point by point subtraction of buffer and sample, then scaled subtracted sample data and averaged them (what CHROMIXS does)?