Generally I think it makes sense (as munenori has already said), that an ISTp-Si subtype (who either is relatively stronger in Si/Se, or values Si/Ne relatively more [than a Te subtype]) would benefit from more Ne (than a Te subtype) and hence be more suited to the ENFp-Ne, and vice versa.

What might be interesting though is how subtypes are different in their usage of (unvalued) id functions. I started a thread recently where thehotelambush suggested that as an INTj-Ne, he finds Te much more annoying than INTj-Tis do, and also said that some ESxj-Si's have more noticeable demonstrative Se.

Now I have read before that creative subtypes (e.g. ISTp-Te, ENFp-Fi), have more prominent PoLRs, so for example an ISTp-Te may be more sensitive (insensitive?) to Fe than an ISTp-Si. But at the same time, an ENFp-Fi may have more noticeable Fe, which might not be a good thing for the ISTp-Te. Perhaps then this ISTp is better off with the ENFp-Ne, because they are less likely to hit the oversensitive PoLR.

However in the same post, thehotelambush mentioned that he found demonstrative Se very attractive, so who knows how this stuff might work.

A related question - is duality likely to be 'stronger' for two accepting subtypes compared to two producing subtypes?