>>1
Your error is in assuming that only 1 dating method is used per sample. Samples of known age (for instance, records indicating the death of an important person) are compared to results gathered from dating methods such as Carbon-14 dating. Scientists predict what the C-14 test will show and after extrapolating the data, the predictions (most of the time) are usually within a few years of the known age. Now, C-14 isn't the only dating method available. There are dozens of dating methods, each with their own range of ages they can check for (for instance, Uranium dating is fine if you want to test a sample that is 500 million years old or more, but it's piss poor in dating a sample that is 1000 years old). Since many of these dating methods have ranges that fall within each other and since each one is built off of other methods that we know to work, and since numerous methods give the same results, then we can be pretty fucking sure that the methods work.
But what if a sample gives 2 different answers for 2 different methods?
1) One of both of the methods employed may have been a bad choice for testing the material (for instance, using C-14 and Uranium on a specimen that has been dead for 500,000 years, which is far outside of the range for both dating methods)
2) The sample may simply be a poor sample that wasn't in a closed system
3) The dating methods simply do not work. However, if this is the case, then there is are 2 huge questions that need to be asked: How could the dating methods work on the other millions of samples of known age, but not on this sample? Also, why did the different methods agree so much before, but not now?
Chances are, the last option is least likely. The first 2 options are the most likely culprits for any discrepancies.