I just took Scott's notebook and added code to integrate the input-referred temperature noise power spectral densities over different bandwidths of frequencies to see what noise we should expect with 16 Hz sampling and what does the measured data suggests. The expected total noise sum is below 1mK for bandwidth up to 10 Hz and the measured total noise is below 0.3 mK up to 10Hz bandwidth. In my opinion, this circuit has passed the test if our experiment and the analysis is correct subject to Rana's opinion.