Apparent 'stuck bit' in ADC readings

[white-ar]

Configuration:
DC voltage into A8
Starting voltage: 0V incrementing in 1.5mV steps.

Between approximately 0.2049V and 0.2197V the ADC reading is 4054 (16 bit resolution) or 252 (12 bit resolution)

The chart above shows the very strange behaviour. Notice the other discontinuities above and below. Sorry for the very poor annotations!

Board firmware version: 4.2.4

Any thought on am mbed interim workaround?

Many thanks

[white-ar]

An additional measurement. Voltage on pin A8 measured using Brymen BM869s (high resolution DVM).
Arduino readings from pin A8 taken as an average of 100 readings taken with 5ms between them.
Readings taken 5 seconds after the voltage changes - to allow the ADC to settle.

[white-ar]

Reading the same voltage scan via mbed directly, using: AnalogIn ain(PC_2C); (I'm not an mbed expert, but this seems to work). The averaged ain reading is: ain * 10000
gives:

[white-ar]

Perhaps this issue is related to the STM32 ADC calibration see https://community.st.com/t5/stm32-mcus-boards-and-hardware/possible-adc1-differential-linearity-error-issue-on-nucleo/m-p/619855

Not sure how this relates to the Arduino environment.