I had the same thought when I watched some documentary about that. Unfortunately I couldn't find any references either. What I did, though, was to start studying medicine, and things started to make sense.
When you exercise there is an acute increase in heart rate. That increase is short lasting, and depends on your oxygen debt (thats what exercise physiologists call it) during and after training. Once that oxygen debt has been repaid, there is a slight increase above baseline as the body is in a higher metabolic state (anabolism to repair and to build muscles, organelles and the like).
When you compare the total number of increased heart beats during exercise and after to the lower number of heart beats throughout the day and night, due to increased fitness and oxygen utilization efficiency, the result is a lower total number of heart beats.
I do not have any studies to back this up. However, you can read about these phenomena in exercise physiology books. For one side of the coin you can compare heart rates of athlete populations and sedentary individuals (I did just that; kind of like a meta-study), and for the other you can yourself measure (with a pulsmeter) your total increased heart beats, from baseline, for a training session, and you will see that it adds up. Furthermore, the last 6 years I have been recording my heart rate several times per day, together with every kilogram I've lifted in the gym, and every meter I have run or sprinted. Having gone through periods of both maximal fitness and injury periods with total rest, I have found the total number of heart beats to indeed be lower the more you exercise (unless you become overtrained as I was for about 3 months). This could be a good study to publish. However, since athletes differ widely in their physiologies and types of training, a case study wouldn't be as good as a prospective study with at least 10 participants.