After calibrating utilities to purchase likelihood scores, the utilities indeed are no longer zero-centered. This is to account for the fact that different respondents have different intercepts to predict whether they have generally higher or lower purchase intent.
I would only use the calibrated to purchase intent CBC scores within the Purchase Likelihood simulation model in SMRT's simulator. If you are planning on running competitive simulations among products using Randomized First Choice, Share of Preference, or First choice methods, I would not use the calibrated to purchase intent utilities. They will be too flat.