As Bruker's OPUS file format is not described openly, we do not know its exact structure. One problem is, given only a few series, how to decide which are absorption spectra? Our solution, empirically developed, is as follows:
1. Remove broken series (such as ones where FXV > LXV, missing NPT information, etc.)
2. Remove interferograms. (See [simplerspec](https://github.com/philipp-baumann/simplerspec).) Interferograms have a starting value of 0.
3. If after these two steps we still have more than one series left, choose the one with highest average value. We empirically verified that other series are usually random noise with values near 0.