Here is the reading of yesterday's brew.
It was a particularly dark stout, BeerSmith said it would end up as 106.4 EBC.
Now, because it is such a dark beer and as we've seen the ability of the software to distinguish darker beers is quite poor (evident from the standard curve), I decided to dilute the target beer to bring it into a useful colour range. Apparently it is standard approach to dilute dark beers before colour measurment.
Here's a screen shot:
Note that I used the original uneditied photo from my iPhone (it came out quite over-exposed), and I added labels in PS afterwards just for clarity.
Looking at the curve produced from the standards, the best looking curve is the "red channel only" so I used that for estimation. Sure enough, the dark end (70 - 100 % on the x axis), is almost flat. This means the software will not pick up any differences in colour for dilutions between 70 - 100 % (i.e. 35 - 50 EBC). It doesn't really recover until below 50% too, so I opted to use the 20% dilution of my target beer for measurement. It came out as 21.9 EBC.
Multiplying 20% back up to 100% gives (21.9 x 5) =
109.5 EBC for this stout, not far off the prediction of 106.4.
This was a measurement of the sweet wort before fermentation, so I'd assume things might be different once fermented and fined (I know, I know, no need to fine stouts, but I do anyway

).
Promising first steps perhaps?
Cheers!