Hello,
No, you are exactly right. Cura will overestimate with a rectangular cross section compared to a circular cross-section.
So, if you are extruding more material with Cura's estimate, something else has to change. That something is the distance between path spacings ( I would guess, but Cura's code in not open source).
If you look at my post on Slic3r versus MC, the path spacing between adjacent extrusions are also different. Slic3r calculates path spacing as a function of both layer height and extrusion width, while in MC it is equal to the nozzle diameter. Both calculations are important when considering flow through the nozzle
But let's think about this: the 'rectangular' and 'circular' shapes for a second.
There is a huge difference between the geometry that the slicer assumes for flow calculations and reality. The slicer simply assumes a geometric cross section to aid in the calculations of the E values. This is a problem inherent in the fact that FDM printers utilize GCode to produce parts. Without this geometric consideration, it would be impossible to predict the E values, which are based on the conservation of volumetric flow rate.
So when it is stated that the filament extruded is 'circular' and 'not rectangular', I cannot see how that is possible. I think the correct answer would be neither because the shape is extremely variable and dependent on a huge number of variables, such as material flow, ambient temperature, nozzle temperature, head speed, layer height, etc. The shape changes as these variables change. From experience, I would say the shape is ovular, and changes with respect to these variables. As layer height is increased, the cross section approaches that of a circle. As layer height decreases, the cross section approaches that of slic3r's cross section.