When you split the loop into two subloops, the restriction properties (from the perspective of the pump) change drastically. That, in addition to the D5 being out of the 'fat' part of its efficiency curve, is what allows this design to work so well

And yes, when you go in parallel, the sum of the flowrates of the two subloops/branches/arms/whatever you choose to call them = the flowrate through the pump.