For those interested, this blog post on training peaks cleared this up for me and confirmed that I was wrong.
That said, I think my (and perhaps others) confusion, in part, was/is related to calling the test a HR drift test when in fact the HR only drifts when the pace is kept constant. When HR is held constant, the pace drifts. Either way it gives the same results, and since, as Scott mentions, its easier to maintain HR (and not pace) when running outdoors this is the easiest way to measure the uncoupling of these two parameters. I may simply call it the drift test going forward.
Off hand, however, I do not know if TP would give a negative value if you hold HR constant, and a positive value if you hold pace constant. If this is true then perhaps this explains why some users receive negative values. In the end its the deviation from zero that denotes drift, or decoupling, and not whether the value is positive or negative.
Thanks Sashi and everyone else for ‘arguing’ with me, sometimes I need a little dose of the Socratic method to clear things up.