It’s a non-compliant MP4.
Its codec type is AVC (which is a subtype of h.264 / MPEG4)
The profile encoded is High@L3.1, which is limited to 3 reference frames at that resolution, but that video is encoded with 5.
… and if they can’t get THAT right, who knows what else is messed up…
http://en.wikipedia.org/wiki/H.264/MPEG-4_AVC suggests maximum reference frames is calculated using the formula:
Min(Floor(MaxDpbMbs / (PicWidthInMbs * FrameHeightInMbs)), 16)
Which equates to:
Min(Floor(18000 / ((720 * 404)/256)), 16)
Which gives a result of 15, not 3. Hence there is no issue having 5 references frames at this resolution encoded at level 3.1.