It looks like I am able to post here again. Don't know what happened. This subject was examined a little about a month ago. You can search for my comments back then but I will state again similar to what I said then. Some transmitters do not like to see any reactive load on them. They will cut back the power out and/or generate some spurs in the presence of a reactive load. Hooking a transmitter to a dummy load or an antenna with a pretty flat swr can present a good load to a transmitter as the load is broad band and so are most antennas fairly broad band. But when you put a cavity/duplexer on to a transmitter it is anything but broad band. It can have a good match/return loss/swr at the wanted frequency but at the same time be very reactive off frequency and the particular transmitter may not like that reactance.
It is not just a matter of tuning the duplexer to present a good return loss at the wanted frequency as is done with a network analyzer or other method of tuning. The transmitter may still not like it for the reasons stated above. As Jeff has stated, the transmitter is not a 50 ohm source but is set up to deliver power into a 50 ohm load. Its output impedance may be far from 50 ohms but it doesn't matter as that won't affect the tuning of the duplexer. As long as the transmitter can deliver power to a 50 ohm load we really don't care what its output impedance is. However what impedance is presented to the transmitter as a load is important and that depends on the cavity tuning and whatever the interconnect cables does to the impedance if the cavity is not presenting a pure 50 ohms. The cable will transform that impedance to something else. This can be on frequency reactance and/or off frequency reactance. There are several ways to correct the problem of a transmitter not liking the load that a duplexer presents. The best and most expensive is to use an isolator at the transmitter. This always presents a flat 50 ohms to the transmitter. It may not transfer the most power if there is some (on frequency) mismatch at the duplexer though. Some of the power as the result of a mismatch at the cavity will go to the load on the isolator and be lost. Then there is the small inherent loss of the isolator and low pass filter too. As a side note here; an isolator can rob you of power into the antenna if the antenna/and or duplexer have reflected power. All the reflected power presented to an isolator goes to the isolator load and is lost in heat. Without an isolator nearly all the reflected power that is present will make it back to the antenna and be radiated as it gets re-reflected when it reaches the transmitter. That is of course if the transmitter is happy working into a reactive load. The other ways to correct the problem, other than using a different transmitter that is not bothered by reactive loads as much, is to use a Z match or try different length cables that make the transmitter happier. Both of these transform the impedance/reactance presented by the duplexer to something more palatable to the transmitter and allow it to produce the power intended. There is no sin involved in using different length cables to make the system work properly. It is not a band-aid approach to mask other problems. The real problem is that some transmitters, because of the way they are designed, do not like reactive loads. Some of the duplexer manufacturers tell you about trying different cable lengths in their tech notes for these very reasons. Close spaced duplexers will be the worst with off frequency reactance as the impedance has to change quickly as you move away from the wanted frequency in order for the duplexer to do its job. 73 Gary K4FMX
