I have never quite thought of it in those terms. For myself, I have always thought that refusing to call the output of GCM predictions intellectually dishonest. The whole purpose for constructing a model and then running it forward in time is to gain insight into the modeled system's future behavior—in other words, a prediction. To call predictions projections and argue that you can not falsify the model based on them is semantic subterfuge and further indication of the fundamental weakness of the IPCC's case.