Mikkel, perhaps you are referring to outcrossing vs backcrossing within the original parents (sweet orange and Poncirus). Backcrossing should narrow the results more than outcrossing would.
The main reason I want to use the segentranges is that they've gone through a very vigorous screening for cold hardiness. The reason I want to use segentrandarins as breeding partners is potential shorter generational cycles, cold hardiness, lower off flavors, and potential earlier ripening of fruit.
In hindsight I suspect US 852 would have been a good seed source. There are a number of tri-specific cold hardy hybrids. These appear to be vigorous in many cases (Thomasville).
I was thinking more of the dichotomy of hybridisations between different citrus species (species hopping), such as Pt x Orange ---> F1 x mandarin --->F2 x pomelo ---> F3 x kumquat ---> and so on
in contrast to backcrossing, where the hybridisations are limited to one parental species.
In theory, I imagine a crossing plan as follows:
F1 cross C35 (Pt x orange)
x selfing
---> F2
mass selection (like you did with the C35 seedlings)
backcross of the hardiest with Citrus
---> F3
then: mass selection for desired traits
backcross to the same crossing partner like in F2
and so on.
Apart from the time it takes...
I would guess that by concentrating on a few traits and limiting it to one backcross partner, the variance of the offspring can be better limited within the desired traits.
The hybridisation of new Citrus species is also promising, but has the disadvantage that the number of seedlings must be larger, because the combination possibilities are simply more divers and there may be undesirable gene combinations that lead to completely different traits than in the parent species. Ilya once gave an example of sweetness, which is encoded by different genes in 2 citrus species, but the combination of both leads to sour hybrids (I hope I remember the example correctly).
Both has its reason and depends on what is to be achieved. I will do both as there are so many other factors that are involved (e.g. no flowers on a certain plant and the year is lost and so on)
but for a designed breeding program I would prefer a backcrossing scheme.
Wild hybridisation is good to find new types and new combinations.