You're speaking of expose to the right, which in oversimplified terms is the practice of always bringing the highlights just short of clipping, and then adjusting exposure in post. The reason this is done is to get the best accuracy out of the analog to digital converter. The conversion is done from a continuous voltage curve, into a 14-bit binary representation. If you underexpose, for the purpose of avoiding ISO noise, you may only be using 9 or 10 of those bits, instead of all 14. ISO is an amplification of the continuous voltage in the analog domain, before the digital capture is performed. Increasing the ISO will increase the noise floor, but so will boosting exposure in post, the difference is that if you boost ISO, you still get to use the full range of the A to D converter, and so any adjustments done in post have more data to work with.
So, does exposing properly throw away dynamic range? It may not change the absolute range.. but if by properly you mean "so it looks right", then in certain cases, it does throw away digital dynamic range, which in the end is more restrictive than throwing away analog dynamic range by boosting ISO. You may only get 12 of the 14 bits utilized in a "proper" exposure, meaning if you want to do work to the shadows in post, you'll see more quantization error than if you had boosted the ISO, even though you may still see the same absolute noise floor.
I'm not sure if I'm explaining myself well.....