[CIRCULATOR] What are main factors affecting to isolation of a Circulator?

Hi all,
I am studying some stuff related to wideband Circulator device in range of 2.3 to 2.5 GHz
and the question is how can i control isolation route (magnitude and phase) ? for example from Port 2 to port 1
For what i ve done, to get insight, I bought a commercial Circulator and do analysis it with Network analyzer, but when i decapped off and solder it onto new and PCB, wierdly It did not provide the idential isolation (phase/magnitude) ( I want the phase here flatten as much as possible and same magnitude response when moving from this PCB to another PCB)
I confusedly do not know exact issues, i guess due to heating or area of ground plane of PCB but i can not say for sure what culprit is

is there anyone that played with it, please give me some advices ?
TM

The specified isolation is under prerequisite of perfect matching 50 ohm matching of absorbing port 3. Any returning port 2 power reflected at port 3 will be passed to port 1. Respectively PCB layout and termination in the test circuit will matter.

Hi FvM, thanks for your clueless
Regardings PCB board, I designed all routes of 50 Ohm Microstripline to mitigate reflection back
However, Isolation from board to board always changes in the range of a couple of dB consequently it exacerbates the overall system performance
Moreover, there is a distortion or inconsistent results measured in the phase that i dont know how to explain (take a look at my phase response below): some get flatten but others get linear

Do you have any expertise on how to control this one ( i wanna this phase flatten as much as possible - I am not a circulator guy then digging into circulator design is such a big deal)
Regards