If you put it like that every EC2 performance metric is ambiguous when not using dedicated instances.
So there are a lot of metrics that are not ambiguous. Imagine if they told you that you get a "moderate" amount of RAM, and maybe it could be 4.5GB, or maybe it's 9GB.
There would be benefits if Amazon gave SLA style promises for these sorts of instances. When they see that they are falling short, they redistribute loads or add hardware to meet the promise. Otherwise it's generally somewhat meaningless because if you can't plan on it, you can't build a plan around it.