The problem is that there is a difference between being confident in a conclusion and being confident that a particular factual proposition is correct. Following the GPS procedures allows you (and more importantly others) to be confident in your conclusion. However, your conclusion might be that a person is "maybe" or "more likely than not" or "probably not" another person's parent. No procedure can guarantee a black and white answer because evidence is often incomplete and inconsistent.
I just read the BCG post about getting rid of "preponderance of the evidence" as the standard that must be met to "prove" kinship:
http://www.bcgcertification.org/skillbuilders/learyevidence.html Having gotten rid of POE, BCG nonetheless says that its standard for "proving" kinship is unchanged and is somewhere between POE and "beyond a reasonable doubt" (which BCG mistates as "beyond a shadow of a doubt"). It says that kinship is "proved" when "the evidence supporting it is of sufficient power to convince a reasonable, unbiased person." Unfortunately, this just kicks the can down the road. Convince a reasonable, unbiased person of what? That it is more likely than not that the person is the parent? Convince a reasonable unbiased person so they don't have a shadow of a doubt that the person is the parent?
Most crucially, however, is the question of whether we need to even make a "yes" or "no" descision at all? Here's where breaking from legal conventions make a LOT of sense. In a civil case, the plaintiff either wins or loses; in a criminal case, the defendent is either guilty or not guilty. There has to be a yes or no answer. In genealogy, however, that does not have to be the case. We should be free to acknowledge that nothing is ever 100% certain and no matter how much research we do and how many sources we cite, we almost always end up with a conclusion that is somewhere along the spectrum from absolutely impossible to 100% certain. This is where Chris seems to be going. He is introducing intermediate statuses betwen "yes, they are the parent" and "no, they are not the parent" to acknowledge that facts frequently leave us with a very well reasoned conclusion that they "are probably the parent" or "its a reasonable possibility that they are the parent'." I think this a big improvement over BCG's approach, which appears to be clinging to a binary "yes" or "no" standard.
Regardless of how many certainty/uncertainty statuses we create, however, we still need to define them. Of particular importance is deciding when it is appropriate to link one person as the parent of another. Regardless of the reality that we may conclude that there are several possible parents, we are currently still stuck with just choosing one. Since we are moving away from a simple yes/no approach, however, I totally agree with Chris that that standard should be well below BCG's binary decision standard.