aleps_86,

When performing the judgement of the two alternatives you've mentioned, I think the first step is to be sure about the model accuracy. In both alternatives, the truncated models have to count properly the displacement from VAL, and I think you've found a conclusion on this subject.
The modeled buried bends must simulate as accurate is possible the complex interaction between the toroidal geometry of bend and soil. That June 2000 Newsletter, available on site, may be the basis of an improved model.
Maybe you have to pay attention also to the pipeline geometry on the "vertical" plane to the above ground end of pipeline.
And, of course, the soil must be accurately considered.

After rechecking the model, maybe you'll see an improvement of the calculated displacements. As you can see there are lots of "maybe" because I cannot guess what is there without having the model.