Non laboratory TMV - help needed

We are using some equipment in our factory to determine whether to accept or reject each individual unit we make. Our regulatory team have determined that the appropriate validation for this equipment is a test method validation - and we have followed the guidance for that. We prepared samples, tested them through the equipment, generated the graphs showing the relationship between the sample attribute being measured and the equipment response, and fitted the best linear line to that data. All in accordance with the guidance.

However, the real best fit for the data is a polynomial curve (y = ax^2 + bx + c) - and using the linear fit is crippling us becuase it is leading to very high levels of waste (false rejects).

I can find no guidance that says regulators will accept a curve response - can anyone please help? Is there guidance / regulatory comment anywhere?