There's a rule of logic that you can't draw specific conclusions from generalities. The efficiency conversion number itself is an approximation, and while it may get into a fairly tight band over time and numbers of people, it's actual variation is greater. So, you're still back to calculations that are only as accurate as the size of the database, or time interval allows.

It's like the old story of the statistician who drowned crossing a river with an average depth of 3'.

Of course everybody has his own approach and/or belief in numbers, but I'm an old school (before computers made complex calculations practical), and prefer to step pack and look at the big picture.

Watch weight over time for gain or loss. Don't fret if it seems to be 3 steps forward and 2 back, you only care about the trend. At the same time track your riding (also in a big picture sense, ie 100miles/week).

After a while you'll get a picture which will tell you if, on balance, the riding you do, however you tend to do it, is burning more or less than your calorie input supports. The rest is common sense and there are plenty of guides about what to eat, when to do it, and so on. Meanwhile, if you decide you need to shave 1,500 calories/per week (for example) you know if you can do it by cutting out deserts, portion control, giving up starches or whatever -- your call.

The results (over time) will prove the estimates, and you'll have to make your own decisions about sustainability.

In a way, this kind of thing is like a pointillist painting. You can focus on each of the dots or you can step back far enough to let the big picture emerge.