I've come up with a procedure to use Incise Flow and no Fill Basins, therefore allowing some endorheic basins to persist. I would like to know if the results are good, and where does it fails short of realism. The procedure goes as follows:

1- First, I generate a heightfield using this equation (with Mathematical -> Calculate Function):
if((lt(pow(pow(sinc(r/5),0.25),10),0.25),pow(pow(sinc(r/5),0.25),10),0.25)
This generates some sort of circular plateau. It's basically a radial gradient capped at 0.25m (to be fair, this is just so I create an Island-like landmass).
WilburWorldNew.png

2- Then, I multiply this heightmap with some fractal noise:
WilburWorldNew.png

3- I then use Mathematical -> Span of 0 to 1 (for some reason, smaller spans seems to generate less straight rivers for me) and do 25 passes of precipiton erosion, with 1% noise per pass.
WilburWorldNew.png
This generates some nice plateaus, though I see their boundaries with the mountains are pretty sharp, and not smooth, because of the small height span I used.

4-Now, I do one more pass of Precipiton Erosion, one single pass with no noise, just to connect the system a little, but not in it's entirety, for I want to preserve some lakes.

5-Then, I use the Incise Flow feature. I like to use an exponent of 0.1 to get more rivers (I'm probably overdoing it, though), and a Pre Blur of 0.75. The result is this:
WilburWorldNew.png

6- Finally, I do another 25 Precipiton Erosion passes, without any noise this time. I end up with the following result:
WilburWorldNew.png

The result seems to preserve the nice-looking river valleys carved by Incise Flow, while allowing some lakes to remain. That big yellowish plateau is an unconnected basin, as can be atested by the Compute Basin Deltas result below:
WilburWorldNew.png

So far, this procedure is the closest I got to have both Incise Flow nice-looking rivers and allowing some endorheic basins to still exist. I would like to know what do you guys think of the result, does it look ok?