Michael Brookes has stated we need to capture the lesser holdings first, which sounds like the smaller station, Friend Enterprise. But he also stated in a Civil War, we'll get the most valuable a set first. Which seems to contradict the first statement.
So I came up with a theory...
In the beginning every system’s stations’ ownerships were more or less randomized to those faction who were living in the system. Therefore some gained one and some gained more stations.
According to numbers in the image, civil war critical became active when faction A had equal influence as faction B. Instant civil war became active when faction A's influence was bigger than the minor factions (factions B, 4 and 5) together and then captured the station. And with a bit of jump in the future where faction A has equal influence as faction 3 which is the controlling faction both should go civil war critical. And later when faction A has more influence than rest of the factions together it’ll capture faction 3’s station and becomes controlling faction.
This theory is supporting Michael Brookes. Invading faction has to capture lesser holdings first which will be done while gaining influence starting from the least influenced station. If a faction controls more than one station, they only lose one which is the most valuable one. This also supports Michael Brookes.
Now that Dukes expanded to HR they gained 9% influence which is a default value for an invading faction. Since then they have raised influence and gained other factions. What happened to these factions when Dukes had equal influence and what happened when Dukes had more influence than lesser factions together? Stations didn’t change ownership, but maybe they should have? This also means that currently it’s impossible to flip ownership of a system because station flipping is bugged.
Do we have data to test my theory?
PS. I checked numbers in the image and did a slight misscalculation. 5th feb A had already more influence than minor factions together. Either it has a one day buffer before the effect becomes active or influence values are calculated as Integers instead of Floats. For non-coders: integer is a rounded number and float has decimals. Then my theory will work.