Slic3r woes

So I'm trying to print Greg's Wade Extruder from thingiverse. Repetier said it wasn't manifold, so I ran it through netfabb and got a fixed version. The first time I printed it, it looked like it was doing fine until it got to a point were it needed (I guess) some support layers, at which point I got stringy messes in the holes. I looked back and saw that I had left "generate support material" off in Slic3r. Oops.

So I turn that on, and go print again. Aside from the fact that now the print takes a lot longer (I'm okay with that and some settings I changed caused most of the additional time), I've been watching it for about three hours and aborted the print finally. It looks like most of the holes that need to be in the extruder, except for a couple, are filled in. Areas that should be hollow are filled in with support material (I think).

On the first print, the machine tried to bridge in mid-air and didn't do a very good job. Part of it is that, for example, you can't bridge a curvilinear surface in mid-air without any kind of support material. But the overall intentions of Slic3r seemed to be good. That was actually the only problem with the print, albeit a big one.

Now, it looks like it's structurally printing okay, but it's filled most of the holes with support material. The errant material is rectilinear grids, and because of that and the fact that clicking "generate support material" and the infill patterns (infill to honeycomb and concentric so I could differentiate) were the only difference between the two prints, I figure that has to be the problem. This isn't stringing between perimeters because Repetier knows from the G-code that it did this, and the offending material is way too well organized.

So the first two images are what it's supposed to look like. The next ones are photos of the actual product which has been aborted at about layer 70/141. The last ones are the Repetier screenshots from the "G-Code Editor" display showing what it's actually tried to do, along with some zoom-ins, aborted at 70/141.

What did I screw up on settings here? Surely this isn't working like it's supposed to.

In Slic3r, the settings are below. The 0 degree setting on the overhang threshold was telling Slic3r to use it's automatic detection. I'm guessing that's broken.

Layers and perimeters:
Layer height: 0.2mm
First layer height: 150%
Perimeters (minimum): 3
Spiral vase: No.
Solid layers: Top: 3 Bottom: 3
Extra perimeters if needed: Yes.
Avoid crossing perimeters: No.
Start perimeters at: Concave points: No. Non-overhang points: No.
Detect thin walls: Yes.
Detect bridging perimeters: Yes
Randomize starting points: Yes.
External perimeters first: No.

Infill density: 0.1
Fill pattern: honeycomb
Top/bottom fill pattern: concentric
Combine infill every: 3 layers
Only infill where needed: No.
Solid infill every: 5 layers
Fill angle: 45
Solid infill threshold area: 70 mm^2

Then on support material:
Generate support material: Yes.
Overhang threshold: 0 degrees
Enforce support for first: 0 layers
Raft layers: 1 layers
Pattern: rectilinear grid
Pattern spacing: 2.5mm
Pattern angle: 0 degrees.
Interface layers: 3 layers
Interface pattern spacing: 0 mm