If UNC beats MD, and if UNC beats FSU, and if Duke beats Clemson, then it is 50/50 at that point (assuming you give UNC a 50% shot at beating Duke, and Duke a 50% shot at beating UNC).
The 5% number came from using kenpom's numbers, and, most importantly, *before* Duke lost to VT, and *before* assuming UNC would beat MD and @ FSU.
I used kenpom's numbers and assumed a win over Clemson and got the 95% (again, before VT, before it looks like UNC will beat MD -- those hadn't happened yet). Sagegrouse simply pegged every game at 50/50 and the math was still 84% Duke would garner at least a tie for the regular season title.