Gain-jump calibration follows the same procedure as above, with one modification. You must set the timerang and calibrate the pre- and post-jump sections of data separately. It is possible that the jump occurred in a program source between calibrators. If you can determine this from inspection of the data, then fine, set the time range boundaries here. However, often the program source is too complicated or weak to show this clearly and the best solution is to flag out all the program source between the two calibrators which show the jump. You then calibrate before and after these scans separately.
The `slap-dash' approach in which you just ignore the gain jump and interpolate through it, may, in fact, not be too bad. As the two-point interpolation approaches the time of the gain jump, the calibration errors will increase, and then decrease as they pass through the jump towards the next calibrator. Therefore, quite a lot of the data will be reasonably well calibrated. You can then take the approach that, if your source is sufficiently strong, you will correct these errors later with self-calibration. Since the best solution is really not very difficult to do, I don't recommend this second approach unless 'your plane back to Iceland leaves in 1 hour and you still haven't written your backup tape. Do it correctly if possible.