The attached larch script (a modification of your script but note: not python, so you may have to add a bunch of '_larch=mylarch' arguments) to show the effects of smoothing with Savitzky-Golay and of interpolating onto a "classic XAFS energy grid".   You might want to play around with this different options.  It definitely seems that none of these methods are systematically over-smoothing the data and reducing the amplitudes of the oscillations -- one more reason to use Larch instead of Ifeffit.  But it also seems like smoothing this data with a Savitzky-Golay fitler is helpful.

Also attached are plots showing k- and R-space generated from the attached script.