It is most likely that you will want the scale factor to be true at the track height rather than at the height of the terrain above the tunnel, so your seed point heights should reflect this. It might involve some estimation if the heights of the track inside the tunnel are not precisely known, but usually an interpolation between end point heights will be sufficient – bear in mind that 5 parts per million of distortion corresponds to a height variation of 32 metres.

The final SnakeGrid report will show a map of the scale factor distortion that is calculated from the terrain height, and so will not precisely reflect the scale factor in the tunnel, but the accompanying statistics on RMS and maximum scale factor distortion will be calculated at the tunnel height.