As the others have said, this is not all that great a discrepancy.
There could be any number of factors in play here.
First, you need to remember that the Census did NOT ask for a year of birth ... it asked for an age. The online sites that provide census information generally subtract the declared age from the year of the census to give a "year of birth". BUT ... most censuses were taken only about a quarter of the way through the year (late march / early April ... although 1841 was not until June ... but adults' ages in the 1841 census are supposed to have been rounded down to a multiple of 5 years, although this didn't always happen). This means that for about three quarters of the population, they were actually born the year BEFORE the year computed in this manner (take me as an example - I was born in September 1967 so the 1971 census will show me as 3 years old, and this manner of computing will give you a birth year of 1968 when in fact I was born the year before).
Secondly, people in the past didn't tend to count their ages obsessively in quite the same way we do. In the days before the Welfare State and automatic qualification for all sorts of benefits at age 60 or 65, nobody really needed to know their ages after they had reached the age of majority (21). It wasn't a relevant piece of information to them. Every ten years, however, the head of their household was asked to declare their ages for the purpose of the census, and he or she did the best they could in the circumstances. Sometimes their estimates could be egregiously inaccurate. In general, I discount ages in the 1841 census (except for under-15s, whose ages are supposed to be precisely stated rather than rounded down) and work on the assumption that apart from this the first census in which a person appears is LIKELY to be the most accurate as to age. There are exceptions, however. If a person is a servant, for instance, then there is a good chance that their age has been filled in without reference to them, and may be inaccurate. The same applies to places of birth for servants ... the employer is far less likely to know for certain than the parent, so prefer the evidence of a census return completed by the parent rather than one completed by the employer.
Thirdly, minors (that is, those under 21) required parental consent to marry. If the parental consent was not forthcoming, then they might go somewhere else to get married and lie about their ages. My great grandparents certainly did. They were both born in 1891, they were married in 1910, and their marriage certificate records both as being 21 years of age. They married in a different registration district from the one where both their families lived. They both gave the same address on the marriage certificate ... and address which is never again encountered in the documentary record relating to either of them, and which looks like an accommodation address where they lived for just sufficiently long to satisfy the residence requirement. It was a register office marriage, and both witnesses were from the bride's family. I somehow think there may have been an issue with the consent of the groom's father ...
I think those are the three main factors that are in play. The end result is that you shouldn't be too worried about minor fluctuations in apparent years of birth between one recorded age and the next. We all have them in our trees.
