A couple of explanations:
The "annualness" of tree rings can be correlated by comparing tree rings with volcanic activities of known dates. You've heard of the "year with no summer" and other such historical events; they look for these in the tree rings as one method of determining whether more than one ring is produced per year. If you find the eruption of 1601 or 1602 accurately portrayed in the rings, you have helped confirm the reliability of the method.
Some trees do produce more than one ring. The primary ones used for the US calibration curve are the bristlecone pines found in the White Mountains of southern California. They are very stable.
And they extend far older than 5,000 years. By matching up ring patterns in living trees with those from standing dead trees they have exceeded 12,000 years.
Other such information comes from oaks in Europe, corals, glacial varves etc.
The daughter product in radiocarbon dating is not important, as it is not measured. What is measured is the residual C14. And the initial amount is what is established by comparison with the tree rings. If you date a series of rings, say, 9,670 years in age you can establish the degree of correction at that point. This takes care of atmospheric fluctuations.
Radiocarbon dating is only used back some 50 or 60,000 years. The millions of years would be other forms of radiometric dating.
Hope this helps.
Religious belief does not constitute scientific evidence, nor does it convey scientific knowledge.