Hi again,
I have checked. Actually MP is using the spherical assumption to calculate distances (to make sure just get distances for points at a same longitude and realize that they are proportional to the angle in radian). The radius seems to be 6378200m .This explains why calcpos actually works well. However, as a consequence distance calculations have a certain error. It's about 15 m by tenth of degrees for mid latitudes which I guess is bearable for most needs (around 0.27%). It would be interesting to know what microsoft has to say about this -not much advertised- approximation in the calculation of distances...