I was trying to wrap my head around the Offset Element Method (found in the Applications Guide under Online Documentation) when a thought occurred to me.

Would it be acceptable to use the Near Point Method and offset the connecting node vertically by the amount of the bend radius? This would keep the dummy leg acting along the centerline of the vertical run and place it in the proper place on the bend curvature, but it would be much easier to calculate and input than R(1-cos alpha).

I have tried a couple test models though and the numbers do not come out very similar at all. Am I missing something?