The strategt I use is simple:
1) Scan 16 or 32 (lots of square corners) scans with the object standing upright. Then I register at the y-axis and 22.5 or 11.25 degrees (you'll need a stepper controlled turntable in order to do that) and use a global fine allign.
2) I then combine the first section of scans into a single object (combine not fuse that is importent). Afterwards I scan a second series of scans with the object on it's side......same procedure.
3) I allign both combined scan using a free allign and contact pair selection
4) I decombine all scans
5) I run a globel fine allign on 36 or 72 scans
6) I allign, if all goes well hura!
7) If not I remove all scans that are not need to close the bottom and top of the object I'm scanning......and cut up the remaining set in order to have a few overlap as possible during fusing (overlap is your freind when alligning but your worst animy while fusing)...
I've got the proces down to a nice pipeline now....so much so that I more and more reduce my scanning to two or three scans of the top and two or three scans of the bottom of the object (or place the camera slightly above the object to cover the top in the first rotation).
BTW, do not overscan....if an object has no surface detail, try don't to add that with insane meshresolution.....that has no use (it only clogs up the rest of the production pipeline, wether you go to print or to CGI).