NavList:

The Sun's lower limb was all that I ever used for decades. Never the upper limb unless that was the only limb showing between clouds. When using the natural horizon as reference the lower limb remains prefered primarily because of the ease of determining tangency when swinging the arc. The moon doesn't seem to offer a choice. What you see is what you get which means having to deal with the upper limb. Superimposing only got done with stars to determine index error. Limb to limb was always used to determine index error by Sun.  A few years ago I started using the artificial horizon after reading many posts here on the subject. After trying different powers of scope and observing upper limb, lower limb, and superimposed Sun images I have found no differences in LOP accuracy. This is odd upon reflection but so be it. The net result is that for artifical horizon observations it is personally more efficient to use a low powered scope and superimpose the Sun image. Limb to limb for index error determination is no doubt the best way but superimposing does save some math and beats a fuzzy horizon. 

Greg Rudzinski