@FLRacingFan explained well what the logic is, but I'm with you. The problem with "analytics" is that it analyzes the game as if two computers were facing off.
A team that maintains a 7-point advantage late in the fourth quarter tends to play quite conservatively. This increases the chance that the other team will get another possession. A team with a 6-point lead that has just had a two point conversion scored against it is going to feel a lot more vulnerable, and a lot less willing to give the ball back. That is the human element that is lost in pure analytics.
I get the idea that scoring two is supposed to increase pressure on the team with the lead, but i don't think it tends to work that way. The very real risk that you could manage to score two TDs and still not even force OT because of two missed conversions makes it a dubious gamble IMO.