Another vote for Photoshop. I frequently do flat stitching from the technical camera through Photoshop using "Reposition" layout, which does not transform any of the source layers/images. When making horizontal panoramas, I always shift the back as far as the camera will allow, then crop after the images are blended. Shifting less just duplicates the overlap and gives you less of the original scene to work with. On a 54x40 sensor and +- 20mm, there is a little less than 14mm of overlap, or 25% of the frame. I find that to be plenty for Photoshop to almost always recognize and apply the correct positioning of the two images. The only exception is some ocean/water scenes where there is both wave action and significant cloud movement between the two images. On the rare occasion Photoshop cannot align the images, it is relatively easy to do it manually since there is only one primary dimension to worry about. I just zoom in to the layer interface and bring one of the layers over in small increments using the keyboard arrow keys.
You have to be a little more careful when horizontal stitching with the back vertical. At 18mm, I find the 4mm of overlap (10% of the sensor) to be enough in a static scene, but Photoshop will struggle if there are moving clouds or water. Again, they can be manually aligned relatively easily. There may be some necessary cloud repair with Content Aware Fill after the images have blended. At 20mm+-, there isn't any overlap in this orientation so you really do have to take a third image on-center. Again, that is on a 54x40 sensor.
Assuming you are using Capture One on the raw images (or Phocus), I apply LCCs and all other global corrections before they are blended. The joint is near the lens optical center, so you can usually get away without LCCs even if there is significant vignetting. Obviously that is lens-dependent.
I often apply a slight graduated mask across the sky before blending if the sky needs to be darkened. However, that mask needs to be copied and aligned across to both images in order to eliminate brightness differences between the two files; you cannot be too heavy handed prior to blending since the skyline is usually not linear across both images.
Dave