Interpolation of collisions unreliable for small N?

[jorindevandevis]

This happens on branch: https://github.com/Wall-Go/WallGo/tree/ReproducePM

If I run StandardModelReproducePM with N = 5 and corresponding collision terms, I get a very different answer from when I run it with N = 5 and N = 11 collision terms. In the latter case, interpolation is used, and this does not seem to work well. The M = 88 GeV point becomes a runaway in the interpolated solution.

For N = 3, all solutions become runaways in the interpolated case, whereas none are runaways with the N = 3 collision files