Misterbenn, thank you for this very interesting paper; my conclusion is to remain on the low side with a shunt resistor. For that purpose I ordered a module with a INA219B chip and a 1% 0.1 ohm 2W resistor on it. I can then also simply read the output voltage through a divider on an input of a microcontroller connected to PSU GND (and connected to the INA219 for current measurements).
I have not yet tested with a stable voltage source (probably this week).
Please let me know where to find the 0.1% tolerance shunt resistors, so far I could not locate them. I have a Fluke 289 that measures to 0.01 ohm but if 0.01% is needed that is not enough. The INA has a 1% 0.1 ohm resistor, so that will be 0.100 ohm which I cannot even measure.
But then my experiments with the ACS712 learned me that the 10 bit ADC I am using implies a resolution of less then 4.8mV, which corresponds to about 10mA (without counting noise, drift, ..). Now the INA has a 9 to 12 bit ADC, so not much better. I start to get it that I may be happy with a display resolution of 0.1A. Therefor I think that the 1% accuracy of the INA shunt should be adequate.
But hey, may reasoning may be wrong somewhere, please correct me.