I don't regard the calorie estimates by my Polar FT7 to be accurate numbers, but I suspect that they're better than you would give them credit for. I've been estimating calories burned on my Kurt Kinetic trainer by finding the average power I produced using their published power curve, and then using the typical formula for converting power into calories (although to use that you need to make an assumption about your own personal efficiency at converting food calories into power at the pedals). I acknowledge that using a calibrated power meter is generally going to more accurate than the Kurt Kinetic power curve, but the power curve number is probably a decent estimate.

I'm finding that the Polar FT7 estimates calories about 25 - 35% higher than my calculated calories. The Polar estimate is higher when my average heart rate is higher during a given session, even if the average power is the same. The Polar appears to be estimating calories per heartbeat, and it doesn't have a way to know that my HR was higher or lower during a session as a result of my average cadence during the session or other variables that can affect HR.

One thing I'm wondering about about: do calories calculated from power output include calories burned during that time due to other bodily processes? As noted above, it's necessary to use an efficiency factor of 20 - 25% when converting power into calories. (i.e., 20 - 25% of the calories went to mechanical power and 75% - 80% was lost as heat, etc.). Say you calculated 500 calories burned per hour based on power output, did your body just burn 500 calories during that hour, or was it 500 calories plus base metabolic rate, which for me is probably around 100 calories per hour for a total of 600. If I was able to add 100 calories per hour to what I calculate from power, and compared that to the Polar estimate, the Polar would still be high but not by a whole lot.