Data Reduction Frequently Asked Questions
How can I run the pipeline when there are no bright objects in the individual image?
The default use of the visir_img_combine recipe of the VISIR pipeline is with the "ref" option set to false. In this mode, the VISIR pipeline does a shift-and-add of the images purely based on the offset information included in the fits file headers, i.e. it does not try to detect the image positions, or expects an input file with explicit anchor points. So for a faint source this is typically how the pipeline is run. Only when the "ref" flag is set to "true" the visir_img_combine recipe will try to cross-correlate the images based on their contents (optionally with user-defined anchor points), which seems to be the situation you are describing in your e-mail. So simply running the VISIR pipeline with the "ref" option set to false should produce the desired results.
You are indeed correct in noticing that in both cases (with the "refine" flag set to false and true), the "visir_img_combine" recipe is trying to detect the sources in your image. However, this step is taken after the basic combination of the nodded images is created in order to create the image with the combination of the beams within the combined image (i.e. the combination of the three (for parallel chop/nod) or four (for a perpendicular chop/nod direction) star images within the combined frame).
The "visir_img_combine_onebeam_0000.fits" file (which contains this combination of the beams), is only created if all beams are either detected automatically (when the "refine" flag is set to "false"), or specified by the user (when the "refine" flag is set to "true" and anchor points are given), which explains why, in your case, it was only created when the "refine" flag was set to "true". So for faint sources it is indeed usual that the "visir_img_combine_onebeam" frame is not created, and one has to derive the source flux (or an upper limit to it) from the "visir_img_combine" stacked image.
So in your particular case, my advice would be to run the "visir_img_combine" recipe with the "refine" flag set to false (i.e. the default setting). The VISIR pipeline will create the combination of the nodded images using the ESO SEQ CUMOFFSETX and ESO SEQ CUMOFFSETY offset information in the fits file headers. If the source is so faint that it is still not seen in the combined image (the visir_img_combine_xxxx.fits file), the combination of the beams will fail and no file with the combination of the beams (visir_img_combine_onebeam_xxxx.fits) will be created. The upper limit to the flux of your source can be derived from the visir_img_combine_xxxx.fits frame.