Seam optimizer gives loops, near-loops, and excessive deviations from an optimal path

See attached 'visualize' file for details.

Seam line optimization seems to be giving closed loops and near-closed loops that visually appear to have much better solutions. I played with the various weightings (anneal, smoothing, optimizer-weights) and although I can pull the seam closer to the NFT, the loops are still problematic. From a 30,000-foot view, it appears that the optimizer should be weighting in the LENGTH of the seam line as well as it's cost and distance from the NFT. The results I'm seeing appear to be the result of minimizing a mix of the distance from the NFT and the average cost of the path (which would be insensitive to the path length). If the length were taken into account, then any loop would be eliminated since it can only add cost to short-circuiting it.