ADC-->DAC latency.
Monitoring one's own voice with headphones is prone to have issues with latencies involved as real-time bone-conduction mixes with delayed acoustic signal. And the air-bourne total acoustic signal (of guitar and voice) reaching the ear is already a mix of real-time leakage signal (even with closed-back) with the delayed playback.

You might want to check again, with actual recording and later playback to take that variable out of the picture.

How does higher samplerate decrease latency?

Good question.

@MC: thanks Matthias, but a couple of things I still do not understand, have a "knot in my brain".

Let's take an interface like the UFX+ without different D/A filters or DSD to simplify / bring it more to the point. From manual:

1. Why is there a difference in converter latency anyway ? Lets take A/D 44.1 vs 88.2 kHz (0.28 vs 0.14ms).
On the one hand the processing intervals are cut at half when doubling the sample rate.
On the other hand you have to process the double amount of data.
So what's the reason why the two things basically don't cancel each other out and lead to a halving of converter latency?

2. Then at quad speed the converter latency is even lower (10 samples) compared single and double speed (12..6 samples).
Is that a property of some converters that they can convert faster/more efficiently at certain speeds?

3. Why is converter latency for D/A often the same at all sample rates ?

4. That this smaller converter latency has a positive impact on RTL (converter + USB latency) you can see in the DAW.
But also there it is somehow astonishing that with higher sample rates the latency decreases more, than you can explain with "lower converter latency", so what is the 2nd reason that RTL decreases additionally?

It would be very kind if you could explain this. Unfortunately, I have not found anything suitable and just do not get it.

Yes, this makes sense, and I should've thought about it! It probably explains most of the differences I hear.

Thanks!