Regardless of how many samples we collect, and how careful we do it, the auto-calibration optimization scripts return non-zero costs, and sometimes calibration values that are obviously wrong.

We want correct, zero-cost output from the calibration scripts. For HP4, this means the buildup compensating version.

With near noiseless simulated data, we see that the algorithms find the correct calibration values, so data noise must be reduced to achieve this milestone.