I think you have the splitter that splits the RF signal from the network analyzer in the wrong place.
Usually you split the signal immediately after the RF Out, so that half of the signal goes to the A-input of the Analyzer, and the other half goes to your controller (here, the laser diode controller). Then you would take the output of your controller and go straight to the actual laser diode, with no splitting in this path.
Here our device under test is the photodiode. So for the reference I wanted to retain the response of the laser diode controller. Otherwise I have to consider the transfer function of that LDC too. I may check both the options at the time of experiment.