Based on the idea that we have to feed the u- / v-offset of the pattern with our path I connected the testmap directly to the U-offset.

Avoiding all u-component stuff and the fractal sum path the outcome was like expected ...

The path is where it belongs by uv-map. U is directed by the map and V comes from the v-component running through the complate group. OK, fine.

Only the pattern's replication is to high cause we omitted completely the factor 80 for the U part.