Hi,
We may never expect the calculated TMR ratio (for Fe/MgO/Fe, several thousand percent) to be comparable to that in experiment (typically several hundred percent for (Co)Fe/MgO/(Co)/Fe). There're also quite a lot of computational work trying to explain the deviation, say, PRL 105, 236801 (2008).
The calculated MR ratio also depends on the model you use, for instance, how many MgO layers there are. TMR would differ a lot w.r.t. different number of MgO layers.
baizq