Because written records were almost all destroyed by 16th-century Spaniards, quite a lot of guesswork surrounds the translation of their calendar to ours, and it appears something went very wrong with the calculations. The Mayas used 4 different calendars, all of different lengths, with the longest of which counting out ages of roughly 5200 years. Figuring out how these relate to 'our' calendars is a big problem, which scientists had thought they had figured out about a century ago. (That's where the 2012 date, which now turns out to be almost 2 centuries out of date, comes from.) However, A German geologist showed in 2005 (in his dissertation) that the proposed correlation to GMT didn't fit with a lot of Mayan-observed events that we know about, and calculated that a roughly 208 year correction was needed, meaning the soonest the Mayan Calendar can end is in 2220.
The final blow was arguably the thesis that nature scientist Andreas Fuls three years ago doctorate at the Technical University Berlin. Fuls pointed out that the GMT-correlation not consistent with a preserved Mayan table on which the positions of Venus are listed. And so there is more, such as inscriptions and objects in time of Goodman, Martinez and Thompson were not detected or outdated. By adding to it all, comes from a very different Fuls dating: one that 208 years has shifted. The end of the long count by the correlation is only about two centuries, at 21, 22 or December 23, 2220. "It is the only option," says Fuls if you ask him about it. (Google translation)
Until then, it would appear we are quite safe, except from Hollywood."