APP's approach of using an alpha channel to remove moving objects from a source image to avoid ghosting when smartblend can't figure things out is an interesting one.

I am just curious if there is a good workflow to do it?   Right now I have to identify the bad images, load them into my photo editor, edit them, save them back out as a tiff or png, go into the layer editor, remove the original image, add the new edited image, and hand connect the new image where the old one used to be connected.   This is a lot of work (Especially compared to the pano program I used to use most, panorama factory, where you just redraw the blend lines with the mouse to avoid the moving objects.)

I tried using "guess location of unlinked images" but didn't have any luck.  Am I using it wrong?

Short of an internal interface for placing the blend zones (which I still think would be a handy thing) there are many ways this could be easier.  It would be nice if one could right-click on the image you want to hand-edit and have it offer loading the image in photoshop or gimp.  Then, after I save out the new image, a right-click on the image to "replace this image" would work well, or some other easier flow.

Or am I just doing it wrong?

I've found a number of other places where, due to either parallax or mistakes by APP in warping photographs that I get discontinuities.  It generally does a good job, but when it makes these mistakes I am keen to find ways to fix them.