I had the same confusion but here is the explanation:
Intuitively speaking, at t = 1 sec , you have 8 Mb of tokens in the bucket, so either you can fill 2 Mb into the bucket and then transmit 10 Mb of full capacity or else transmit 8 Mb which is not the full capacity.
But the first choice is made, the reason being that the the bucket is not full currently, you want to transmit at full capacity and currently you have 2 Mb less than the full capacity and hence you will prefer to first fill the bucket completely and then transmit. Hence 2 Mb is first put in which takes 1 second and then at t = 2 sec, you transmit the complete 10 Mb.