Remove the need for C2 and correct C1 value.

[s3c]

The notes say that it is uncertain why C2 is necessary, this is because the value of C1 is incorrect, in fact it doesn't match the value used in the link provided. An AC simulation of the circuit gives:

This shows that the input signal is sitting on the shoulder of the high pass filter. Removing C2 and using 200nF (as given in the link) for C1 gives us the correct response for the frequency range we need, it no longer sits on the shoulder. We now have a signal that has uniform attenuation across the band of interest, and the decreased attenuation will make it easier for the phone to pick up

For my design I used 10uF for C1 (with C2 removed), which gives us the full audible range from 20Hz upwards. As for R4, the Android spec says the line should be loaded between 32 and 300 Ohms, dunno what's up with that, but here we are.

To comply with the spec, I use a 270/30 divider here (from phone to radio), with another 10uF cap to the output, just to be safe. This combination gives us about the same frequency response as for the input side. I haven't investigated this fully though. The cap probably isn't needed but let's do DC block just to be safe.

Note that the 10k on the output of this divider is just to simulate the radio input impedance.

[kronicd]

That looks great!

To address your point on R4, that was my addition as detection on Android devices was not working correctly.

The related discussion is here #14 and the value was chosen due to this issue elsewhere and the specs here. I note these don't match (but do overlap with) the specs you've linked. It is indeed frustrating.

Have you tested your revised design with Android and iOS devices? If you can do so I'd be happy to contribute an updated schematic and pcb layout.

If you need some PCBs for testing I'd be happy to send you some and you can swap components as needed. My email is my_git_username (at) my_git_username (dot) net if you want to reach out.