Real world model with data quality

Hi all,
I am trying to use what we learned in class to apply on a real world problem. I have 2.5 years worth of 5 minute data. I have a model that could do some ok prediction but would like to improve it. I found out that my numeric features has an associated data quality boolean : True for good and False for bad. This boolean is an indicator to let us know when we have telemetry problem for that particular analog feature. Are there any example on how one would incorporate data quality in a model? Should we throw away all segments where data quality is bad?

Thank you.

Based on your post, seems like a datapoint with bad data quality should be ignored. Do you have a strong reason to include incorrect data as model input?