The ECU filters that data to extract knock signatures and assigns any detected knock activity to specific cylinders according to firing order and crank / cam position info.
The ECU filter must filter out road surface induced noise and other vibrations using a road roughness factor and other non combustion related noise from the raw knock sensor signals. If the knock sensor/s are loose, or if there is some unusual vibration noise effecting the signal, that can be mistaken by the ECU as combustion related noise. Logging the raw knock sensor signal voltages isn't accurate for knock detection, but the raw data logged may reveal some erratic signal noise that provokes inappropriate knock related timing corrections.
However, your log of the timing retard for knock correction does not seem involved, as the knock correction graph doesn't follow the variation of the ignition timing graph.

Try logging other parameters until something is found that shows variations with a pattern similar to the variations occurring with the ignition timing at the same time.

BTW, how do you know the throttle angle was 100% during the time interval logged? The DBW eGas throttle angle is varied by the ECU, and isn't necessarily only according to the accelerator pedal position.