The sound of any headphone (including an IEM) is partially determined by the way the wearer's body acoustically interacts with the device, and is not purely a function of the device itself. This is why frequency response curves for the same headphone taken from different test rigs often look different. This means not only that the recording from a microphone can vary from what you hear, but that different people (with different ear structures) will often hear different things from the same headphones.
For instance check out this measurement from Harman comparing headphone frequency responses on different people and test microphones:
Another factor: in the video, the IEM is mounted to what appears to be an IEC60318 inner ear simulator rather than directly to a microphone. This simulator is designed to mimic the frequency response of a human ear for more accurate measurements at the "eardrum" microphone; however, if you're listening to the recording on headphones, your actual ear will also have this response... in effect, the recording playback as perceived by the listener will contain the ear resonance twice, which affects the treble in large and unpredictable ways.
In other words, the listening demo is a good way to compare headphones, but not a good way to know in absolute terms how a headphone sounds. For instance, if you listen to a demo of a headphone you're familiar with, it shouldn't sound like it does in real life, but you can it as a reference point to better determine how a headphone you haven't listened to yet might sound, relative to that familiar headphone.