I’m not sure there’s a firm rule on that. My understanding is that deviations are for circumstances where the mission of iNat to connect people with nature is furthered by having the iNat taxonomy diverge in some limited way from the patchwork of taxonomic authorities that have been adopted. One of those reasons might certainly be that the iNat community finds it beneficial to accept a taxonomic treatment that’s currently at odds with POWO. Another reason might be that that the iNat community decides that including some name in widespread use provides substantially more value to iNat users than suppressing it because it fails under some provision of the relevant taxonomic code.
I would certainly agree that the latter circumstance should be invoked very sparingly—what’s the point of having consistent naming rules if you don’t stick to them? But I think this is one case where continuing the taxon framework deviation makes much more sense than submerging all reliably identified observations of these Iris cultivars into an undifferentiated higher taxon.
The one other case I’ve seen of this sort was a decision to use a valid hybrid formula that has been widely used in many years of research over an also valid but virtually unused nothospecies name that had been resurrected after about 100 years. These scenarios are distinct, but both favor retaining names that can connect iNat users to available literature on the plants they have observed over alternatives that provide no benefit to iNat users.
On another point, it seems you’re suggesting that I. × hybrida might have the same exact same parentage as I. × germanica. I think you may have missed this comment from @yuri_pirogov on the flag:
The name Iris x hybrida that we use on iNat is Iris x hybrida hort., not Iris x hybrida Retz. The name Iris x hybrida hort. is widely used in gardening literature to designate artificial hybrids of various species of Section Iris , although it may be a late homonym of the name Iris x hybrida Retz. (Research of the tall bearded iris varieties Iris x hybrida hort in the Stavropol Botanical Garden) The name Iris × conglomerata N. C. Henderson, although has been proposed (see FNA), is obviously still Nomen nudum.
To quote the Royal Horticultural Society:
Bearded iris were commonly diploid (i.e. with two sets of chromosomes per plant) prior to the twentieth century when a few irises spontaneously doubled the number of chromosomes and became tetraploid (i.e. with four sets of chromosomes). Iris fanciers in the early part of the twentieth century set out to artificially increase the number of chromosomes to four sets since the naturally occurring ones had a number of useful qualities – notably improved size, substance and colour intensity. By crossing and inter-crossing the irises with the increased number of chromosomes, a whole new race of more robust colourful hybrids were produced. The tetraploid bearded irises became the norm and it is now rare to find bearded irises which are diploid.
Iris × hybrida hort. refers precisely to tetraploid Bearded Iris cultivars within section Iris, which are distinct from the diploid I. × germanica. I’m fine with moving these to Iris × conglomerata if that’s clearer (admittedly still not a valid name under the ICN).
Given that these cultivars are the result of 130+ years of plant breeding, it’s not realistic to expect that the parentage of any substantial number of them can be determined (unlike I. × germanica). Having a single iNat name to apply to them serves the interests of observers and identifiers with minimal deviation from the iNat taxonomy. It reflects the fact that humans have created organisms distinct from those in nature and that many people (especially newer iNat users) will observe these and want to know about them. It does that without burdening the iNat taxonomy with multiple cultivars. And it avoids mixing both cultivars and yet-to-be-determined pure species in an undifferentiated pool of section-level identifications.
As far as I’m aware, it appears that all iNat users who work with these taxa appear to support maintaining this deviation. I do understand the principled taxonomic opposition, but I believe iNat’s taxonomy exists to serve the overall aim of connecting people with nature, and this is a small and reasonable deviation to serve that aim. I’m also confident that there’s a long list of other taxonomic fixes that we can all work on before we need to shoehorn Bearded Irises into POWO’s framework.