A perfect cut is impossible.
A saw is straight, and makes a straight cut. Collars are not straight, and even if they were, the nature of a cut, opens a tree up to pathogens, so what's most important is the size of the cut, tool sanitation, tree health, and things like that.
Lets talk about what a good margin of safety is when making a cut. How far from the collar should we be?
We train our grunts to stay about 1/8" away for each 1" of diameter.
Do you believe by increasing or reducing that distance we could affect tree health? I don't think it would matter one bit if we doubled or even tripled it! I'm sure by reducing it, we'd be cutting into more collars.
You mentioned the stub being food for pathogens. As you know, the branch collar and the trunk collar form a cone shaped structure in the tree where the branch is attached. When you kill a branch by cutting it, there is an area of wood that is abandoned by the tree. Previous to removal, this area is walled off chemically and mechanically to prevent decay, should the limb die, or in this case, be killed.
It is also an inter-twined mesh of the two types of wood, trunk and branch, that gives it another type of decay resistance.
The tree knows that limbs may die, so the barriers are set in place as a survival tactic, if you will. This cone shaped, ingeinious design has the strongest decay barrier the tree has.
It will not matter if you leave a 1/16" stub, a 1/8" stub, or even a 1" stub while cutting.
So no, I'm not kidding.
For to make my point, here are the top 5 google images of proper pruning cuts:
http://www.littleyorkplantation.com/facts/collar%20cut.jpg
http://www.bugwood.org/gfcbook/figure159.gif
http://www.umext.maine.edu/onlinepubs/htmpubs/images/branch.gif
http://www.igin.com/Resources/shigotree24.GIF
http://comenius.gvm.cz/physics/P6210037.jpg