Flux linkage must be determined from the magnetic vector potential
difference
which is an order less accurate than energy computed from variational FEM.
Therefore I expect some differences in the results per se.
>It should be possible as well to model the transformer in the section at
>right angles to the normal one to get some idea of the end-winding leakage.
>Because the leakage is mostly in air superposition theory should apply
>reasonably well and you may be able to just add the leakage fluxes
>together. You need to be aware though that the core material will be at a
>different level of saturation in the "end-winding" model although you
>could make the core material linear and set its permeability to a typical
>value from the "normal" model.
>
>Remember as well that with ferromagnetic or ferrimagnetic core materials
>the inductances will be functions of current because of saturation,
>although this may not affect the leakage too much.
As I see it the affect of nonlinear core material on leakage inductance
can merely be neglected,
at least at the level of accuracy I need here.