When you design a diplexing filter from a prototype lowpass and highpass filter, it's important to get the terminating resistances correct. I see from your symmetrical filter designs that they are matched to finite impedances on both ports.

If you design a (for instance) lowpass filter with (say) 50ohms impedance on each port, then as you swing the input frequency over a wide range, you will see the filter input impedance change. In the passband, it will be 50ohms (+/- a bit). In the stopband, it will tend to zero or infinity, depending on whether the filter is L or C input.

Now in a diplexing filter, this doesn't happen at the common port. This stays matched regardless of frequency. At low frequencies, the lowpass filter matches it ( as long as the highpass is series capacitor input). At high frequencies, the highpass filter matches it (as long as the lowpass filter is series inductor input). That means the voltage on that port __stays constant__ as the frequency changes.  

So the common port is not conventionally 50ohm matched. In fact, you need to design your component filters to match, at the common port, into a short circuit. This is the correct model that will work with constant voltage regardless of frequency. (I had trouble getting my head round this at first)

Most filter design packages will give you the option of specifying the port match. Design the component filters for zero ohms and a series component on the common port, and 50 ohms on the isolated port.

Let us know how you get on.

By the way 'doesn't work' is a rubbish thing to say, it will get you down-votes. The circuit did something, say what it did. You were expecting x, you measured x, it worked. You were expecting x, you measured y, it didn't work, so what were x and y? What was the inserting loss at the cross-over frequency, and far from it? What were the port matches at the cross-over frequency, and far from it? And what were you expecting?