I don't think there's a general answer to this -- the forms you cite appear to be individual random exceptions to the rule.
The long ā of the fut. ppl. stātūrus is historically secondary. The pf. ppl. stătus has short ă, which is etymologically expected (it's the reflex of an Indo-European laryngeal, like in dătus); stātūrus must therefore be an analogical form, based on the long ā that appears in practically all first-conjugation verbs. The noun stătūra, not being a verb form, presumably escaped this analogy.
As for par(i)tūra: there seems to be no noun of either form in Lewis and Short. A PackHum search for partūra finds only one example, in Varro; the normal fut. ppl. form seems to be paritūrus. This fut. ppl. is slightly irregular in that it does not share the stem of the pf. ppl. partus; the latter presumably results from vowel syncope between coronal consonants, which is a common phenomenon (though not a wholly regular sound change) in Latin. In this case the fut. ppl., maybe because it's a less common form, evaded the sound change and remained regular (or was analogically reformed).
Figūra is the only really anomalous form in that it obviously cannot be derived from the stem of the fut. ppl.: in other words the suffix here is -ūr- rather than -tūr-. It seems to be the only such example; Weiss's historical grammar of Latin lists it specifically as such in the section about derived nouns in -tūra- (I don't have the book handy at the moment so can't cite the page).