Anyone who has taken a course in Stats knows that raw statistical data is utterly useless without proper statistical analysis. For example, your only metrics for proving that Mists of Pandaria was a costly mistake, primarily due to the theme and story, were 1) that it had a subscriber decline of 2.6 million and 2) that it has the lowest total number of subscribers to any expansion to date.
The problem with these metrics is that you are missing a metric ton of variables. One of the most important variables missing is why people quit in the first place (Was it the theme? The Pandaren? The new Talent revamp?). This one is especially interesting because it actually has a negative correlation with the theme (Pandaria and the Pandaren) being the cause of the expansion performing poorly. MoP sold 2.3 million boxes, not counting digital downloads, in the first week, and over the first quarter actually gained 900,000 subscribers (Up from Cata's low of 9.1 to MoP's peek at 10 million). Second, by saying that MoP was a failure due to it's expansion low of 7.4 million, you ignore that Cataclysm was responsible for the loss of 2 million subscribers that never resubbed for MoP at all (The difference between WotLK's high of 12 million and MoP's high of 10 million). Without any kind of data knowing why these people didn't resub, we have to simply blame that loss on Cataclysm.
Moving on, even if we say that MoP performed poorly strictly by losing subscribers, we have to assume that this was due to the longevity of the expansion rather than the initial theme. We know nearly 1 million people returned at the outset of MoP, so the theme must have been enough to entice people to at least try it out. Why didn't they stay? There are a long list of variables that could be the cause, none of which are shown in the data we have. Those variables include, but aren't limited to, not liking the other theme of the expansion - the Faction War, the endless amount of dailies, the switch from dungeoneering for casual players to LFR for casual players, general burnout that occurs every expansion, and game fatigue - the permanent burnout that happens in a 9 year old MMO.
Finally, even after showcasing all of this, we know that, thus far, MoP has had more staying power than Cataclysm due to the amount of subscribers each expansion lost. MoP sits at 2.6 million subscribers lost while Cataclysm lost 2.9 million. Granted, this could change with the Q1 2014 report, but there seems to be evidence that the subscriber loss has begun to plateau; the last quarterly report only indicated a loss of 100,000, which is rather low considering it was riding on the weakest content patch of the expansion, 5.3.
Ultimately, saying MoP 'flopped' when it sold millions of copies and still managed to retain a number of subscribers that eclipses all other MMOs in history based on a simple downward slope on a graph is completely misrepresenting the issue. Warlords of Draenor will attract people again, and then lose those same people and more. This isn't because people don't like Draenor, the crazy time-travel plot, Orcs or Draenei, but rather, because the game is aging and had a very clear peak in Wrath of the Lich King. It's all downhill from here, even if they announce a fan-favorite theme for an expansion, such as the Burning Legion.
Mists of Pandaria was a controversial expansion, to be sure (I mean, damn, just look at this thread), but there is no available data to accurately explain the losses without taking a sample survey of the people who quit. The only obvious one is that the game is old and past its prime, and that everyone should expect the losses to either plateau or continue to decline based on that one fact, especially after the initial jump in subs after each expansion's release.