Ribarm,
Just a consideration...
Your resolution to average the z factor seems to be flawed in my my opinion. I know that it is the premise of the current Triang approach, but the fact that data can reflect a difference to a 2d resolution is a challenge to that theory. If 2d coordinates show a zero relationship, it should imply that an elevation difference is desired.
If the TIN is vertical it should not be assumed that it is wrong.
I have used a different program, EZYsurf, to build the TIN's and it succeeded. It took a very long time to create the contours, but it also succeeded. It is using a 3d resolution.
On a 2d evaluation it might appear that the contours cross, but they do not on a 3d evaluation.
I think it is an issue that needs to be examined. It is a tough one, but it is also the reality of contours.
The drawing that is submitted may be poor, and even contrived, yet it is also a situation that can be encountered. Why not attempt to resolve the issue without compromising the data?
Rick