In this dialog is calculated what theoretically the outcome would be if a model is made according the chosen levels of a factor. Normally only significant factors are used in this calculation. When the model is build according the confirmation run and the result is not according the calculation it could be that a interaction is missing or something was wrong during measuring/building of the experiment.
In the Level confirmation run column the level of the factor can be selected to be used in the confirmation run.
When selecting the level of an interaction the level of the factors in the interaction may not be selected, because this is already done with the interaction.