Is there any evidence that their in-house model values behavior different from FICO's in some fundamental way? For example, that it likes it when you have several accounts with a positive balance, or that it likes a 7% utilization better than a 2%, or whatever?
If not, then as a practical matter there may not be a lot of importance to how it works. I.e. if you want a better score, make sure you have a CC reporting a balance, keep your reported utilization low, have an open loan (painless and virtually cost-free with the SSL technique), pay your bills on time, and work to get all derogs removed from your reports. That may not result in a perfect NFCU score but I imagine it will result in a score that is good enough to guarantee best rates on a mortgage or auto loan (say) or for approval on any of their credit cards.
On the other hand, maybe NFCU does work much differently from FICO 8 (say). If so, it seems like we'd need want to investigate in great detail the corner cases, i.e. where someone claims to have a very good FICO 8 score and a terrible NFCU score (or vice-versa). I can imagine possible explanations. E.g. the person has a clean strong profile except for a recent $70 collection -- that could result in a high FICO 8 and a bad score in another model. But without a vast amount of reliable detail about their reports it would be difficult to speculate.
