I think it makes sense to use a color closer to the infrared we're interested in capturing... If you think of a rainbow color-scheme; the infrared lies just beyond that indigo color at the one end of that scheme... 
Black is just close, and cheap/easy; not the optimal color... 
Edit:
Talking about one percent variations; i'd guess that a random dark indigo color would get you that 1% closer to optimal, compared to black. Could be fun to experiment...
Infrared is at the red end of the spectrum, hence the name.
Ultraviolet is at the indigo end, and Im sure that if there was a better color than black for absorbing heat the panel manufacturers would be using it.
Also: Since you're consuming it as heat, absorbing light at ANY frequency is a net gain. So you want to grab it all. The name of that color (and texture) is "flat black".
Surface coloring controls frequency selective COUPLING between light and surface temperature. At a frequency where the surface absorbs well it will radiate well when it gets hot enough to radiate significantly at that frequency. If it reflects, it also won't radiate as easily. You want to grab and hold as much energy as possible. So selective coloring would only be useful for inhibiting re-emission (which would be mostly in the far infrared).
But you don't need to do that: The glass cover does it for you by reflecting re-emitted infrared back to be re-absorbed (and also keeps the hot air in so you don't lose much by conduction, either). So you can paint it black and also catch what far infrared does make it through the glass.
That's what the greenhouse effect is: Transparency for most frequencies by the glass/atmosphere, downconversion to far infrared of absorbed and re-emitted radiation at the surface, and reflection (or backscattering) of the far infrared by the glass/atmosphere, keeping most of the re-emitted radiation around to be re-absorbed. You gain far more by holding the downcoverted energy captured from higher frequencies than you lose from bouncing some incoming far infrared.
Selective coatings that have low emissivity in the long IR are used on some commercial collectors. They do show a significant performance gain over black painted absorbers. For example, you can see this in the SRCC results for the Heliodyne Gobi flat plate collector by comparing the black painted absorber version with the black chrome selective finish version:
Blat painted:
http://securedb.fsec.ucf.edu/srcc/coll_detail?srcc_id=1981085DSelective black chrome:
http://securedb.fsec.ucf.edu/srcc/coll_detail?srcc_id=2007026DI believe that these panels are the same except for the selective coating on one and black paint on the other.
This link lists all the SRCC tested panels -- I think that Radco also offers black paint and selective coating in an otherwise identical collector.
On the efficiency curves, the intercept values are nearly equal, but the slope on the selective one is -0.837 BTU/hr-Ft^2-F vs -1.07 for the black painted absorber.
If you look at the sunny day heat captured for the category D operation, the selective one delivers 24.3K BTU/day, and the black painted one delivers 20.9 K BTU/day -- so the selective one collects about 5% more heat than the black paint one.
I think your logic is basically OK, but I think that when the absorber radiates heavily in the long IR (as the black paint one will), the glass does block a lot of that radiation, but it absorbs it and gets warmer and loses more heat because of that.
Whatever the reason, selective coatings do show an advantage in every test that I have seen.
If you are building your own panel, the flat black paint does fine, but if you want to get part of the way to what a selective coating will do, the Thermalox partly selective paint that is available in spray cans and has an long IR emissivity of about 0.5 when applied at the right thickness to well cleaned bare aluminum might be a good choice. The fact that good DIY collectors can be built for about $5 a sqft where commercial collectors are more like $30 a sqft plus expensive shipping still favors the DIY collectors on a BTU per buck basis by a very wide margin.
Gary
