The same two characters — P70 — mean three different things in this field, and confusing them produces answers that look broken. Given one accession:
Relative to an empirical baseline: the latest observed year where germination is still ≥ 70% of the accession's early-window best. Nonparametric, censored by the monitoring schedule, defined for every accession that has a baseline. This is the catalog's definition.
Absolute: the fitted curve's crossing of 70% viability of all seeds. Model-based, smooth — and undefined for any lot whose fitted starting viability is already below 70%, which is common in wild collections. When this P70 comes back blank, that blank is the correct answer to the wrong question.
Relative to fitted Ki: the curve's crossing of 70% of its own fitted initial viability. Model-based and defined for every lot — but it's a different quantity from either of the others.
None is wrong. They answer different questions with different tools, and the only sin is comparing one flavor against another without saying so. The toy below computes all three on the same data.