My guess would be that the difference occurs primarily because the MFI is measuring the amount of fluorescence on each individual cell "event", whereas the percentage of cells moving over is looking at the whole population. 

So looking at your graphs, it may be that a minority population have very high expression of the construct in response to treatment, which increases the MFI, while the overall percentage of cells expressing the construct does not show a similar trend.

This in consideration, there is still a significant shift between the two. How are you labelling the construct, and how are you treating the cells afterwards? Also, are you applying gating parameters? It could be that you are acquiring aggregates which are not part of your target population...  Also, what cell population are you using? DCs for example are quite heterogeneous and so you have some intrinsic variability in treatment? Hope this helps!

Thanks Brittany! 

Cells are T-cells, and, of course this is after fine gating, that includes viability and CD3 population.

I did think you would - but sometimes we (or at least I!) just need someone to point out the obvious! Is your graph n1? And are your samples normalised to baseline expression? I am quite surprised that there is so much difference but perhaps further replicates might reduce this also! 

Thanks, Jean-Marie.

Well, negative % is due to the background subtraction...