I agree with your statement, but only if you consider the source to be the music itself - the album and artist = not how it's transmitted to you.
But yeah, if not lossless, the data compression algorithm can cause the source to lose dynamic range, depending on many variables.
However, if the dynamic range is not even there on the album to begin with, caveat emptor - 24 bit 192 kHz cannot give a poor dynamic range CD anything more than what it started with. The DR site below gives pretty clear examples how bad some albums can be.
Here's a site, for instance, that ranks the "DR" by artist / album to give you an idea of what's bad, and good. Go to the FAQ page for more info. Note they don't rank data compressed music at all, as depending on the algorithm, it can a wide range of numbers, none truly representing the original recording as released by the studio.
Dynamic Range of Albums
Also, Foobar 2000 has a DR Meter plug-in component if you want to measure your own stuff.
The guy in the video above has no idea, IMHO, what he's talking about by coming to the conclusions he makes about what we "need" Other than the general statement, that I agree with, that with data rates, uncompressed is preferred. But compressed at 320 kbps (4:1) is way better than 128kbps, and that better than 96 or 48 kbps - all typical rates you can get on internet streams from "free" internet stations.