Yeah series should be better, I asked the same question a while back and got the same answer, but I think I understand why a bit better now.
Think about it this way, going to guestimate some numbers as examples:
SERIES
Lets just guess you get 1.5 GPM if you ran them in series, your overall system flow rate is 1.5 GPM and each block is getting 1.5 GPM, very simple and each block consistently gets the same flow rate regardless.
PARALLEL
OK, so if you got 1.5 in series, removing the one block and splitting out the flow across two blocks will cut the system flow rate in half across each block (assuming they were the same), so you should have less "System" pressure drop. The result of this is a net increase in system flow rate which sounds good, but the reality of it is... you might increase your system flow rate maybe 30% at most because of the pump/system curve relationship. Lets just guess this turns out to be 2.0 GPM system flow rate, sound good right? Well remember that you split your system flow rate across each block, so your blocks are actually only seeing 1.0 GPM.
So series will still produce more flow rate across the blocks where you need it the most, your radiator performance may be slightly increased at the 2.0 GPM, but it's not so good at the water blocks. Then there is that nasty little "water will find the path of least resistance" issue. So... the additional possibility and likelyhood is that your CPU block is more restrictive than the GPU block, in the end you might get a split similar 0.7GPM across the CPU and 1.3GPM across the GPU maybe worse depending on the blocks.
As everyone noted, running in series is the better option both because your system flow rate equals the water block flow rate (not cut in half), and you ensure there isn't an imbalance between blocks.