Hi! Thanks for your interest!
First off, the numbers are an estimate, assuming that the measured accuracies are close to the thruth. It takes reviews from the last two months to make the numbers average, but recently relevant.
Secondly, it assumes that “lessons per day” stays constant (the number you input) during the whole “lifecycle” of the reviews (from learned to burnt). So it’s a bit like assuming that you did exactly 10 per day for a duration equal to the estimated “Average burn time” which is at the bottom.
So all in all, even when working perfectly, it’s just an estimate using a very theoretical model. But it does give an order of magnitude.
Additionally, I admit I have no idea how the resets will affect this 
That said, since it only uses data from the past two months, it should not be a factor anymore after two months passed since the reset. How long ago was yours?