In an old book of Prescott (died in 1859), The conquest of Mexico, the author described Aztec calendar with a year of 365 days. To compensate for the difference, every 52 years an additional 12.5 days were added (so that every other 52-years period started at noon, and the rest at midnight). This gives the average length of the year 365.24038 day which is only 2 minutes short, and is more precise than any European calendar before the Gregorian one. He refers to the research by Gama, Descripcion historica y chronologica de las Dos Piedras, (Mexico 1832).
However I was unable to find a confirmation of this information in modern books, for example, in Thurston, Early astronomy, Springer, 1994. Modern sources such as Wikipedia, Thurston and others say that Aztec calendar is the same as Maya calendar, and none of them addresses the intercalary days and length of the year. Thurston cites some modern works which try to estimate the average length of the year in Maya calendar and concludes that the claims on high accuracy are unfounded. But nobody mentions the ingenious intercalation described by Prescott and Gama. Does this mean that Gama's research was somehow disproved or is it simply forgotten?
EDIT. Following the suggestion in the comments I looked in the book by M. Coe, The Maya. It indeed contains a sentence:
We know that none of the Maya intercalated days on leap years or the like, as we do, and it has been shown that more sophisticated corrections thought to have been made by them are a figment of the imagination.
No explanation, no references. He even does not mention Gama, the founding father of Maya studies. The question remains how the Maya did agriculture with a calendar not tied to seasons.