FWIW, the stat structure in Linux does not include birth time [1]. It only gives you:
atime: The time of last access.
mtime: The time of last modification.
ctime: The time of the last change to the inode.
I assume the stat command is using a filesystem-specific method to get the birth time.
Anyway, I don’t think any of these stats is guaranteed to be consistent with the rest (or even correct). For example, it is common to disable atime tracking to improve I/O performance.
Assuming the data is accurate, I think the other comment about the file being a copy is the best explanation.