Rick said that physiological configurations may have its own contribution in relationships.
For example, I'm LII-5, and I had a SEE-7 boss for some time. That dude was pretty cool and friendly, such LII-SEE couple would usually only have information exchange problems (this includes mechanical behaviours). But I don't think I'd get along with a SEE-8 as easily as this.
ESE's are usually E2 or E3. I think I'm more attracted to the E2 variant, because Twos are more friendly.
I get along with SEI-9's usually without major problems, but SEI-4's are harder to get along with.
I find ILI-5's quite difficult to socialise with, but less than ILI-4's.
and so on.
I have some hypothesis : variants of a type are driven to other specific variants of their socionic dual.
For example, let's analyse the LIE-ESI couple :
LIE's can be : 1 3 5 6 7 8
ESI's can be : 1 2 3 6 7 8
Sorting them from the most frequent variant to the least frequent (very approximately) :
LIE : 3 6 5 7 8 1
ESI : 6 1 2 3 8 7
If this order is correct, I'd imagine LIE-3 would be driven to ESI-6, LIE-6 to ESI-1, LIE-5 to ESI-2, and so on.
Then let's analyse the EIE-LSI couple :
From most to least frequent
EIE : 2 6 7 3 4 1
LSI : 8 1 3 7
There are less variants for LSI. I imagine EIE-2 would be driven to LSI-8, because they are the most frequent variants of each socionic type. Then EIE-6 to LSI-1, and I won't say anything for other Beta Process Dyad for now.
A last example for ESE-LII :
ESE : 2 7 3
LII : 5 1 6
I didn't decompose Etypes to their subtypes because it'd be more complicated then. I think Intimate Fives are driven to Intimate Twos, Social Fives to Social Threes, or such.
But I still don't know the rules well for physiological dualisation.