thanks for the quick answer and link!
I use Mac M2pro. Driver Type is CoreAudio
its a very powerful workstation which can handle big cpu pressure so I don't think there is any external plugin or factor causing the cpu to load latency...

my setup is:
Sample rate 48000
Buffer size 64 samples
I/ Latency 2.10 ms
O/ Latency 2.21 ms
Overall Latency 4.31ms

After reading in the link that rhe sample rate could improve the latency in recording, I tried to record at 96000 / 64 sample and metronome and it seems that it improved the latency a little... but it seems a lot to record at this rate when I will be loading many tracks, plug ins ecc....

Converter Latency is much lower compared to the latency over USB/FW/TB. Nowadays, near zero.
Look at the manual, there is no merit in using double+ speed to optimize latency only for a tenth of a millisecond, right?

Using a lower ASIO buffer size reduces latency much more, see my table in the posting above.

Regarding the use of double/quad speed, you should be aware of the following:

• If I remember right, there are some VSTi which do not support double/quad speed because sample libs are missing for that speed.

• You put more load on your system (double amount of data to process everywhere).

• If you are working with external devices which are connected via ADAT, you have only half of the port capacity at double speed and only a quarter if you worked with quad speed (which is IMHO nonsense anyway). The port multiplexing does not happen with analog ports on your system or those which are 2ch based anyway (AES, SPDIF).

Tip: Check / Optimize your monitoring whether you route over DAW. Better route locally on the recording interface.

To show you how low latencies are, I made the following calculation comparing UFX+ to UFX III.

When playing music life on stage, you have to deal with longer latencies.
9 samples difference in converter latency is only 6.7 cm (audio transmitted through the air)
20 samples difference in converter latency is only 15.4 cm (audio transmitted through the air)

Yep, when playing to a backing track you start playing as soon as you hear it.
The latency for recording, what you play then is known to the DAW by the (ASIO | core Audio) driver.
It reports latency for A/D and D/A.
So the DAW can compensate.

For the submix of your headphones you need to take care to get
a) the backing track from the PC (via software playbacks, middle row)
b) your played instrument, e.g. a piano, directly from the HW inputs in near-real time
The actual recording of the played piano happens sample correct in the DAW due to latency compensation.

True, but focus is on routing, same for Core Audio ...

This is happening when recording a Djambe part. you can see the  main transients on 2 and 3 are out of sync.
I even tried to record from an empty prioject, only 1 channel with metronome but still same problem.

In Ableton I have delya compansation and reduced latency when monitoring activated

I am trying to understand the direct monitoring in total mix to have zero latency. My knowledge of RME's software routing are still limited so I appreciate your patient in guiding me through this process!

When recording in ableton live, I switch off the monitor on the channel that I am using and I turn up the faders of that specific channel in my total mix "first row" so I can listen what is coming in the mic directly in total mix. Is this called direct monitoring?

Also, the second row in total mix, I can only adjust the volume of channel 1/2 which is the playback software.

Yes.

But you not only have the two mics on the Phones, there is also two faders from the software playback channels up.
Pull down those fully and you will be able to listen the two microphones in your headphone with zero latency.
Without any other signal coming from any software with latency, as Oli told you.

It is not possible for me to record an instrument in a song without listening what is coming out from ableton live. The solution must be somewhere else...

Did anyone ever experienced latency issue due to usb to C cable? A friend suggested me to try change the cable or use an usb HUB, or another port of the laptop as apparently Apple Silicons have had similar issues with other audiocards... I will try this