First, "Importance" is probably not the best way to present the results about the impact of an attribute. And, especially with alternative-specific attributes, it is a bit strange to be comparing the impact of an attribute that only is applicable to cars to another that is only applicable to buses.
Now, I suppose you could focus only on the attributes applicable to buses, and re-percentage their importances to sum to 100%. Then, for buses, you have the relative importances of attributes. Then, do the same for cars, etc.
Still, I am not a big fan of using Importance scores to show results of conjoint analysis. I think it is better to show "base case" results and then modifications to base case.