As far as census ages go, I would think that the 1851 & 61 ages would be accurate as his parents are the ones giving the information, and they match up with his baptism year.

As Sue says, the age at death was given by his son, who may not have really known exactly, or just quoted what his father had told him.

In the 1891 - 1911 there is a huge variation in age, as well as compared to the register. John himself may not have known his precise age, perhaps just that he was older or younger than siblings - it was before the time of birth certificates. Or my current favourite theory, he may have lopped off a few years to minimise the rather large difference in age with his wife - about 12 years if he was born in 1843. The latter I have come across more than once.