Well, if it was a true three month average then each bad day would count about one 90th of the total, or slightly more than 1%. However, BG scores in the month before the test, count more than two months before the test, and much more than 3 months before the test.
I always have something sticky and bad for me straight after my HbA1C, because I know that it won't count towards the next test.
It stands to reason that if you eat a sticky bun (or a chocolate hobnob) 121 days before your next HbA1c test, and the average life of your red blood cells is 120 days. No matter how much glycated hemoglobin binds it itself to your blood cells, the chances are a very small proportion of those cells are going to live long enough to influence your HbA1c result (because most of them will be dead).
This is true, even if you are not an average person, and your blood cells live for, say, 130 days on average. Most of them will still be dead by your test date.
However, if you eat your bun / hobnob a week before your test, then most (but not all) of your cells will be alive seven days later and will contribute to your test result.
So you can see that the further away you get from your test result, the less influence your dietry mishaps have on your HbA1c test.
That kind of random dying of blood cells can be modelled as an exponential decay. The 70:20:10 thing is just a crude representation of the surviving red blood cells over the 3 month period.
My geeky side finds this interesting. But my "knackered eyes" diabetic side realises there's absolutely no point trying to be "good" for a couple of weeks to get a good HbA1c. It's not like you win prizes! It's there to help you see how well you're managing your diabetes, which you kind of need to do all the time so you don't end up being one-eyed for 18 months while you wait for your vitrectomy. Or worse, obviously...