If you have a directional coupler, you are most of the way to an SWR
meter...

Especially if you are willing to make directional couplers for specific
narrow bands (say 10% or 20% bandwidth), it's pretty easy to do. If
you get the freeware "RFSim99" software, you'll find in it a tool to
design several different types of couplers. One thing they don't make
clear is that the coupling for the microstrip and stripline versions is
for a structure 1/4 wave long. The coupling decreases as the frequency
goes up or down from that length. It's certainly useable at half the
design frequency, and in fact you can find articles on the web about
how the coupling varies with frequency, and for structures that give
broader bandwidth. If you only care about ratios (e.g. for SWR
calculation), then the exact coupling doesn't matter anyway. For your
application, at only 20mW, you'll want fairly high coupling. 20dB
coupling would give you 100mV out to a 50 ohm load, for 20mW in.

I've seen articles in "Wireless World" for amplified RF probes. But a
simple diode detector probe feeding a high impedance voltmeter will be
a high impedance load to the circuit you are probing, so long as the
diode has low capacitance. You should be able to find Schottky diodes
that have 0.5pF capacitance at very low junction reverse voltage, I
think, though they are likely to be pretty delicate. (Don't zap them
with static!) With a high-resolution DVM to monitor the output, or
with a DC preamp that has high input impedance and extremely low input
offset voltage (such as the Harris chopper-stabilized CMOS part), you
can see RF voltages down to below a millivolt fairly easily. You need
to calibrate the amplitude response, however, and it drifts some with
temperature. With a buffer amplifier, you could use one of the RF
power monitor ICs from vendors like Analog Devices or Linear Technology
and have more stable calibration, reading "linearly in dB". An
advantage of diode detection is that it's very cheap, and you can
afford to put detectors on your directional coupler outputs and just
leave them there--plus they don't require power supplies. That 20dB
coupler with diode detectors and high resolution metering would allow
you to still "see" the reverse-port coupled power down 40dB below 20mW
input with good diode detectors, and it's very unlikely you'll be able
to make the coupler with such good directionality anyway.

If you don't already have them, you should add "step attenuators" to
your list of equipment. It can be useful for a lot of things. You may
be able to pick up a pair of 1dB step and 10dB step attenuators from
some place like ebay, maybe even at a reasonable price.

Are you going to be making so many measurements that you need a SINAD
meter?? Do you have an RF generator with calibrated output?

Lessee...20mW at 50 ohms is just about 1 Vrms. For that much power,
you need sensitive amplifiers and high impedances?? If you really do
need that for probing, you'll need an amplifier at the probe tip. 1pF
at 1GHz is about 160 ohms reactance. But for that very reason, 1GHz
circuits are designed at low impedances, and you shouldn't need a
particularly high impedance to probe for frequencies, at least.

Cheers,
Tom