Even if a laptop power brick has a three-prong IEC input plug (and many do), that just means the power brick is grounded. It feeds DC to the laptop, usually something like 19V, with no earth ground. 19V is low enough that code does not require protective insulation, as would be the case with 120V (or higher). The power brick having or not having an earth ground means nothing.
The shell of a USB connector is DC-grounded to the chassis of the laptop, but that ground may not be at the same potential as the shell of the USB connection on the other end. If the device receiving power from the USB connection is also getting a different ground reference from some other source, you'll have a ground loop that will move current. But even without that, a floating DC ground connected at the ADC end can pick up common-mode noise from stray EMF in the environment. Digital signals filter that out completely if they remain bit-perfect. But I can see how that noise could find its way into the PCB ground of an ADC, causing noise leakage into the analog circuits.
I have a USB-powered Presonus interface, and also an independently powered Benchmark ADC1-USB. I have never tried to evaluate noise on these, but the Presonus loops back to no better than 100 dB SINAD when powered by a laptop. That is, of course, abundantly good for the things I'm using that sort of device to do (such as live recording and even some vintage equipment testing with REW). The ADC is not designed to a price point the way the Presonus was, of course, and even though it's much older it's also much better, by probably 10-15 dB. Which is largely inconsequential. But it does have much better controls, though the Presonus has phantom power and sufficient gain for condenser microphones, while the Benchmark is made for line-level inputs.
Rick "suspecting that Benchmark completely isolates the digital USB interface, even to the extent of reclocking it" Denney
