This is a VERY simplified explanation. I'm sure much more detail could be added to it, but I think it is adequate for a high level explanation.
The image you see on the screen in your raw editor already is a bit mapped conversion. Your raw conversion program displays the the image according to the default values built into it. The extra information is there in the raw file, but not displayed on screen. If the conversion no longer needs certain data to meet your needs, it is deleted from the on-screen representation, but not from the raw file itself.
As you make changes, your raw editor goes into the raw file and displays any additional information it needs to create the image as you want to see it. But at no time does the image you see on screen represent the ENTIRE set of data contained in the raw file. I suspect we would not recognize it as an image at all if it did.
As an ex DBA, I tend to think of raw images as a sort of mini database for the creation of a single image; not as the image itself. As you work with the image you read that database looking for the information you need to create the image.
Once you have altered the on-screen representation to your satisfaction, and you tell it to export or save to a tif file, what gets converted to tif is that onscreen representation. And this leaves behind any unused data in the raw file when the tif is created. And you would not see any loss in a good conversion compared to the onscreen representation.
This is why I claim that any conversion from raw to tiff is inherently 'lossy'. The raw development process is one of deciding WHAT raw data to use and what to leave behind.
It is also why you always want to save your raw image files if you can. You never know when an alternate development and conversion will be useful.