Suppose that the MtGOX thief has 500'000 BTC divided into 150'000 addresses (3.33 BTC per address on average).  Each day he issues 75'000 transactions (one every 1.15 seconds).  Each transaction combines two of those parcels and sends them out to two new addresses.

He divides the addresses into two groups, A and B, 75'000 addresses each.  The two groups are alternated, so that for the first 12 hours each day the two inputs are chosen randomly from the current A group, and the outputs are added to the new A group, until the current A group is exhausted.  In the next 12 hours he does the same with the B group.  Thus the bitcoins that are input to each transaction are at least 12h (80 blocks) old.

Assuming that each transaction pays a fee of 0.00013 BTC, here is what he contributes:

  Number of transactions per day: 75'000

  Total transaction output BTC per day: 500'000

  New addresses used per day: 150'000

  Total fees (in BTC) paid per day: 10

These numbers match the current blockchain statistics.  Note that 10 BTC/day of fees is a small price to pay for tumbling the MtGOX stolen coins.  

Is this correct?  The point is that the current blockchain traffic is small enough to be "faked" by a small group, even a single person.

Anyway, those blockchain.info plots are rather strange.  They don't show increasing adoption.  I don't know what they are showing.