Whilst everything Teno says is true, quite a common scenario is a USB charger with captive cable - in which case the designer does control the cable resistance. And if you don't have a captive cable then remote sensing isn't an option anyway.
Entirely true about not being able to rely on much more than 3.3v + LDO drop-out voltage.
This isn't a derogatory comment about this tip but one shouldn't think they are getting accurate voltages to the load when remote sensing is not used, even with this tip.
This method only correctly compensates for static cable losses when the cable and contact resistance are known and constant. Unfortunately, especially in USB applications, the cable resistance is typically not constant because different length cables with varying quality can be interchanged. Furthermore, USB applications are often used in mobile applications where the cable is flexed and the contacts have varying strain and therefore varying resistance.
This point doesn't critisize the tip or make it any less valid. It should serve as a warning to USB device designers to avoid pushing the designs to the limit because USB has very poor power regulation because it doesn't use remote sensing. I would design devices powered by USB to run from a 3.3V or less power supply regulated off the 5V USB voltage or if 5V is required use a buck/boost regulator to keep the voltage at 5V. Don't depend on having 5V at the end of a USB cable.