Every longevity statement is a statement about time — so define time once. The pipeline's clock is YearsDifference: the whole-year age of the accession when a germination test began, computed from collection date to test-start date over 365.25, rounded half-up.
Whole years are a deliberate coarsening. Monitoring tests cluster around anniversaries; integer years make "the latest year that met the threshold" a well-defined maximum instead of a smear of near-duplicates. The lab's curve fitters, by contrast, run in days — finer time is better for fitting a continuous decline. Both are valid; the sin is mixing them without converting.
Watch the two classic clock bugs: seeds that sat un-banked between collection and storage (the clock starts at collection, not accessioning — usually right, since ageing doesn't wait), and Excel serial dates imported as the number 41432 instead of a date.