Disclaimer: I'm an experimentalist and not a theorist. 
 
I'm not sure that Zhao and Montano's calculations are correct.  I think their calculation for the transition probability assumes that the electron density is constant when, in actuality, there is a spatial dependence resulting from quantum confinement effects.  As a result, the energy of the surface plasmon peak as calculated by Zhao and Montano is constant for all cluster sizes at approximately the classical value of omega(bulk plasmon)/3^0.5 and shows an increase with decreasing particle size.  Including quantum confinement results in a blue shift of the surface plasmon resonance with decreasing particle size and a decrease in transition probability for decreasing particle sizes (as suggested by Anatoly).  This change in the calculation will also impact the bulk plasmon and I don't know what that trend will be.
 
As far as their experimental demonstration with Fe nanoparticles is concerned, I really can't make an assessment about the experiment since they don't indicate how the nanoparticle size was determined and I'm not familiar with their preparation of the nanoparticles.
 
I'm not sure from this paper what the impact on the MFP will be although there very well could be some effect on the EXAFS.  Personally, when I began working on EXAFS of nanoparticles and small clusters (Au13), I was leery of using the bulk phase and amplitudes to fit the data.  My experience to this point has been that there have not been observable deviations of coordination numbers for nanoparticle systems using these FEFF inputs.  That being said, I tend to agree with Matt that this is something that people working in this area need to think about a bit more than it seems they have judging by the literature.  I would like to see some (correct) theoretical assessments that support the assertion that bulk phases and amplitudes are ok for nanoparticles that I think has been demonstrated by validating EXAFS results with other techniques (mostly electron microscopy).
 
-Laurent