There must be some kind of mathematically-based explanation, but basically, the front sight is closer to the target, so if it is kept in alignment with the target spot, varying the rear sight causes a certain degree of misalignment and resulting loss of accuracy. If the rear sight is kept on the target spot and the front sight is varied, the degree of misalignment is quite a bit more.
Sight dimensions do play into this also - wider rear notch with front sight pushed off to the side will allow more misalignment than a narrower rear notch.
I just got through doing this exact drill in class a little over a week ago (using the alignment scheme Tom described.) We did it at 5, 10, and 15 yards. For me, at 5 yards I had smaller than a palmspread on the target, and at 15 yards I had approximately a handspread, maybe a tad less.
Technical excellence supports tactical preparedness
Lord of the Food Court
http://www.gabewhitetraining.com
On the contrary, at times I've felt that demoes of this with the front sight centered on the target was a bit of a parlour trick.
In general, I've found that when my sights are misaligned, it is the front sight that tends to get misaligned, not the rear sight. Possible reasons include the rear sight being closer to the point of articulation (wrists), and the fact that the rear sight is larger and thus is more visually dominant in centering the sights on the target, whereas the smaller front sight has a tendency to play a "fine tuning" role.
I think it's worth doing it both ways, and every other way one can imagine. Get as much first hand data how one's own gun, hands, sights, eyes, etc. behave and correlate to hits on paper. Doing it only a specific way to ensure getting an expected outcome is a waste of time and just a dishonest pledge to some sort of dogmatic ritual. But yea, it's important to know the subtleties of what exactly it is that you're trying.
FWIW, one time I got around to trying the variation where you keep the target centered in the rear notch and maximally misalign the front sight. I have some pics in my journal somewhere. For me and the guns/sights I tried, I was surprised to find the shots still closer together than I had expected at 7 yards. Also, I found it was actually a bit challenging to intentionally set the sights up this way, suggesting that my natural misalignment tendencies are probably somewhere between the two extremes. Unsurprisingly, it taught me the same lesson it is intended to teach: that trigger control reigns supreme, first and foremost, and that the next most important thing is the relationship between the front sight and the target.
Ultimately, the semi/subconscious SWYNTS and FWYNTF (feel what you need to feel on the trigger) can only be developed through extensive live fire practice, shooting many different targets at different distances and in different physical circumstances. I've found the drill we are talking about to be useful from the standpoint of creating a foundational intellectual understanding of a basic and important condition of aiming - front sight somewhere, anywhere, within the rear notch. Knowing how much accuracy we can get out of THAT particular level of sight picture is a good stepping stone along the way to a finer, less conscious, and more useful-at-speed understanding.
Technical excellence supports tactical preparedness
Lord of the Food Court
http://www.gabewhitetraining.com
Agreed.
What I was trying to say is that I think the variation of the demo where you keep the front sight perfectly aligned with the target but only shift the rear sight around isn't really an honest representation of the sort of misalignment that will occur when one willingly accepts a sight picture with the "front sight somewhere, anywhere, within the rear notch", at least not for me.
As you've noted, however, that variation is the most forgiving when it comes to how good or bad the hits on paper end up, and that's why I have sometimes felt a certain "parlor trick" quality to it. Not to say it's bad, and I probably came off hostile, which I didn't intend.
I think a good lesson to learn from it is that, regardless of how one tends to use or misalign the sights, one should really pay attention to the relationship between the front sight, specifically, and the target. I think one of my problems is that I have a tendency to pay more attention to some sort of general relationship between the target and the total sight picture, and then the relationship between the front and rear sight separately. That's just what comes most naturally.