There are two ways to answer "how long do these seeds last": report the latest test that still passed (observation), or fit a curve and read off a crossing (model). The observed interval's superpower is that it cannot overreach — it never claims more than someone actually saw. Its cost is censoring: the answer is clipped to your monitoring schedule.
Censoring isn't a nuisance to hide; it's information to notate. If the series crosses below the threshold between years a and b, the true interval is bracketed: "a–b". If the last test ever run still passed at year a, the interval is right-censored: "a+" — at least a, ceiling unknown. If no test ever met the threshold, all you have is an upper bound: "<b".
Read "18+" as an ongoing success story (nobody has watched long enough to see the end) and "18–20" as a completed measurement. A catalog that wrote both as "18" would be quietly lying twice.