When extracting a subject from one image to another using the Quick Selection Tool and Refine Edge:
1. When is it better to make a selection of the background, make a layer maks of it, invert it, load the mask and use Refine Edge VS. just making a selection of the subject and using Refine Edge?
You'd never do it quite that way, because you wouldn't add the layer mask untill you had optimized the selection, but it is certainly a valid technique to select the bit you don't want and invert it prior to using Refine Edge. This might be with a landscape where the sky offers the flatest colour and tonal range, but you actually want the foregound selected.
You would do it that way when the object being extracted is on a white background. You would place the replacement background on top of that image. Set the background replacement layer's blending mode to multiply for the lower white background will not effect anything on the replacement background layer. You would then add the layer mask containing the silhouette of the lower layer object and unlink the mask. Your inverted selection. The mask does not need to be very good for any fringe areas cut off by the mask would still show because of the multiply blending. Unlinking the mask let to move the replacement around.
Replacement background on top set to multiply
Add layer mask and unlink it
unlinking the mask gives you the ability to move and position the replacement background