Deviation is either standard deviation or standard error of the mean I guess (abbreviated STD and SEM usually).
You can google them to learn more about them but basically it tells you how much you sugat varies. Basically:
If you have two readings such as 4 and 7, your average will be 5,5.
If you have two readings such as 5 and 6, your average will be 5,5 also. But you vary less and thus your deviation (from the mean) is less.
This is an important variable because big roller coaster values are awful for your health, for instance if you spike massively after a meal and get hypos during the night.
Regarding the hba1c, it estimates it based on the 90 days average. But I think overall it will underestimate your real value by say 5 points about, because people don't test their blood 1h after the meal, and usually that's when it's the highest. Therefore your post meal BS is not taken into account in the estimate, which biases it slightly toward the lows. The less you spike after a meal the less this should be true however.
I have had only 1 test since I started using this app, and it was estimating it 4 points lower than reality. However I have been changing my diet a lot (trying to destroy those stupid spikes) and my estimate a month before the test was the exact same as the real value I got. I would like to see what value it gives me when I have a stable lifestyle over three months.
I think it is great overall nonetheless, because if it goes down it means you are doing things right. If it goes up it should trigger you alarm before nasty habits settle in. So it is actually very useful to me.