Intended knock retard affecting throttle response?

Hi all,

Using COBB on Mazdaspeed3 (MPS), cruising conditions after engine has warmed up would show a constant high count of knock retard. From what I understand, this is intended OEM behaviour to save fuel (advance ignition to limit at low-mid load).

I have been experiencing an inconsistent response to my throttle (how power comes on) and I believe it is related to how quickly the ECU allows the KR to "decay" under load.

Example scenarios:

1. Stab the throttle from low RPM (High delta TPS), car takes off like a rocket, no KR reported.

2. Cruising where KR is active (low load), stab the throttle, very poor response and sometimes minor buckle.

2a. Cruising KR is active (low load), let off throttle (0% TPS, KR instantly disappears), Stab throttle, Car takes off like a rocket.

Key to mention here: KR under full load was the second thing I tuned for (from the OTS map) after a MAF calibration.

Any suggestions appreciated! :D

Update:

Logged good/bad conditions back to back and it looks like the numbers don't lie.

Looking under a specific load (@WOT) and RPM (~4.5k) the ignition timing is about 3deg. But when KR is active @ 3 counts, it is at about -1deg. In addition AFR reflects the WOT (Knocking) map.

In any case, it is pretty annoying and I can't imagine this is how the vehicle was tuned from factory, so I will continue tinkering.