What to do when one Ci is estimated as 0?

[KarinSchork]

In rare cases, Cis are estimated as exactly 0, which leads to an error term of NaN. This subsequently leads to problems in the code for protein elimination, as the error term is compared to the reference one + tolerance.

example: D5, A_B, 1st graph (the problem only occurs on linux, so maybe the underlying optimization gives slightly different results?)