Looks like the rupture ran to the south, with the bulk of the moment release where Chen Ji says it is. Notice the red patch breaks about a minute after the initial patch. The rupture travels slowly, about 1km/s, between the two asperities, usually it would travel 2-2.5 km/s with full-speed rupture.

I wouldn't put much credence in the details - notice the slip runs right up to the edge of the rectangular model both updip and downdip of the red patch, so either the square is too small or resolution is not so good. Also, those little arrows show dominantly strike-slip motion in the yellow patch under the red patch, unlikely, subduction faults generally show nearly purely thrust motion. The scale "depth" is a little funny, the fault is more horizontal than vertical (dip = 27 degrees), so when the depth changes by 1 km, one has moved 2 km along the fault plane.

At this point in the Sumatra earthquake analysis, estimates were that the rupture was less than 500km long, rather than the true 1200km, but the methodology has improved since then, and many of the problems were due to the vast dimensions of that rupture.