When talking about estimated attenuation we often say that adding speciality malts will reducing attenuation. But this is seldom reflected in estimation calculations, is there a way of working out what the effect of adding more speciality malt will be on the attenuation?

I don't worry about the effect of malts used on attenuation. Attenuation is affected FAR more by the strain of yeast and the mash temperature and time. Yes, if you use a very high proportion of crystal malts, you could reduce attenuation a bit. But under normal circumstances, I wouldn't be concerned about it. Or are you just trying to understand the science of things? You could run some experiments. I run a lot of experiments myself, but this isn't one I have been too interested in (and probably not too many other brewers, either).

You would apply estimated attenuation rates to a grain bill/mashing regimen you are familiar with. I get around 80% ADF from my Munich Helles.

1.048 - 80% ADF = FG of 1.010

If I were to substitute 10% caramunich and assumed it was 100% unfermentable

(48*0.9)-80% ADF = 8.6 plus the points from the unfermentable caramunich 48*0.10 = 4.8 gives an estimated FG of 1.013 for ADF of 73%.

Where would the percentage of unfermentable sugars come from for each type of grain?

Most of the software that is out there already calculate for this as much as they can, they take the average fermentability of a grain given to them by the manufacturer and it is accounted for in the software. However, as stated there are so many other things that go into attenuation that the grain portion is a very small factor, including temp, yeast cell count, and time as mentioned above.

OK, that's great information but leads on to....
Which number on the data sheet that comes with grain gives the average fermentability?

Thanks everyone for the answers so far.