Garmin was never right so I just quit updating it. I don't look at what it reports for calories (until someone asks like in this thread).

I found and kept this from someplace on the Al Gore Web and it comes close enough for me.

Multiply the average power in W by the time in seconds to get Joules; divide by the kcal/kJ conversion of 4.184; then divide by your metabolic efficiency for cycling (generally in the range 20-25%) to figure out the number of food calories you burned to get from A to B (in addition to your base metabolic burn).

Assuming a cycling power efficiency of 25% and looking at a recent flat ride I did the calculated numbers came out to 1606 versus a Golden Cheetah/Training Peaks reported 1683 (1686 reported by Strava on 20 May.

If I use the 'barely got into the hilly ride category' I did last Sunday the calculated numbers are 4416 versus a Golden Cheetah/Training Peaks reported 4641 (4654 reported by Strava on 25 May). I'm not gonna fret over a 200 calorie difference in a 6,000 calorie day.

So, as the kids these days would say "EpicFail" for HR based numbers... imho

Roger that.

I plug my data into Training Peaks and it does all the calculating for me