Once intervals exist per storage condition, they can be compared — and the comparison that matters most in practice is 5 °C versus −18 °C, because a minority of taxa are damaged by freezing and a bank that defaults everything to the freezer will quietly destroy them.
The strong signal is the ratio: if the frozen interval is at most half the refrigerated one (R18/R5 ≤ 0.5), something about freezing is hurting the seeds. The weak signal is asymmetric confidence: 5 °C has statistical support at its latest interval while −18 °C has none, and the raw ordering agrees (R5 > R18) — suggestive, so it requires both conditions.
Missing data means no verdict (NA), with one principled exception: a species whose demonstrated interval comes entirely from −18 °C data has already answered the question — seeds that survived twenty years frozen are not freeze-sensitive, flag FALSE. And every flag should carry its reason in words; a boolean that can't explain itself can't be audited.