With thanks to all for their comments. I've given this much thought, and on the basis of advice from everyone who has contributed, and some modelling using Martin King's software, I'm not for now going to tamper with the IMFs, but instead try to duplicate the performance in a different shape...