There is no way to tell APT where to put an auto-wavecal. It will always try to do it just before the first observation at that wavelength setting and then again at regular intervals. Sometimes it is possible to craft the program so that some auto-wavecals naturally fall into occultation, but the only way to move more wavecals into occultation is to turn off the auto-wavecal (WAVECAL=NO) and put in equivalent lamp exposures (target name WAVE).
This has a number of disadvantages. First, turning off the auto-wavecal for one exposure turns it off for all exposures in that visit. Second, the GO-specified wavecal is not associated with the science exposure, so the pipeline can not use it. The pipeline will thus not produce extracted spectra (x1d files for the MAMA detectors, sx1 files for the CCD detector) or rectified spectral images (x2d for MAMA, sx2 for CCD). Instead, the observer, and any archival users, will have to select the proper WAVE exposure and reduce the data themselves with CALSTIS.
WAVECAL=NO is an available mode, so it can be used only by request. Since it makes the data more difficult to use for archival purposes, we strongly discourage GOs from turning off the auto-wavecals just to save a few minutes for science exposure times. We generally require the GO to convince us that the science goals really require using this mode. Gaining an extra few percent in the S/N is not sufficient reason.