First we take a look at the `online' calibrations, a simple subset of what has been proposed by Anders Holm. By looking at some example queries we discover that even the simplest model has the bad query problem. Then we look at a more complicated validity interval model for `offline' calibrations, and see the same bad query problem. Then we show how the `online' calibrations could be handled by the `offline' calibration model and at the very end propose a solution for the bad queries problem. At the very end I show how to handle the case of calibrations changing without correspondance to runs.