In order to obtain a good wavelength solution with CRIRES, you should have at least two or three ThAr lines visible on each detector, preferably well spaced, at least for a 2nd order polynomial. Otherwise the results of the solution will likely not be trustworthy. Obviously this is not a problem of the pipeline but with the wavelength calibration source itself. If you have a few lines then the best you can probably do is to use the recipe crires_model_fix with a model file and (x,y,wavelength) position of known lines.