Fascinating that the reported flight track appear to follow the FIR boundaries. Would be great to overlay the track on your FIR map. The U turn when a/c touched the Indonesian FIR could be consistent with evasion. This would make sense if pilot thought he might have a tail/intercept/monitored after his crossing of the peninsula. If there was a low altitude over the Peninsula and Straits it would also be consistent with radar evasion. I think we have to give evasion a high probability given the known/likely data points
Try
www.skyvector.com -- it has global high altitude charts and a flight planning function that enables you to compute leg distances. Input WMKK-ZBAA into the base flight track, then drag the track line over the asserted waypoints. Bearings and leg distances will be computed and you'll be able to see how they match up with the FIR boundaries.
ETA: Let's see if this works. Link to map here:
http://skyvector.com/?ll=5.687414242...VAL:F.VO.IGREX
Tried to post an image, but the link works. Times are based on an assumed 462 knots i.e. M= .80.