Casa2
I thought the emissivity of a dark roof is higher than that of a light roof. What makes you think otherwise?
It seems to vary depending on which reference you read - however there is not a significant difference between the emissivity of white and black paint.
http://www.infrared-thermography.com/material-1.htm
http://www.monarchserver.com/TableofEmissivity.pdf
I also use to think that white paint had a lower emissivity than black paint. It is the difference between reflectivity and emissivity. The property of high reflectivity means a material reflects radiation well. The property of high emissivity means a material will loose its heat quickly by radiating heat - so if it is hot it will radiate its heat quickly (to cooler surfaces). White paint has a good reflectivity - hence good for cool roofs reflecting solar radiation) but its emissivity is still high - it will also loose its heat via radiation quickly.
Your point about the winter benefit of a darker roof is not one I have really thought about previously but logically if the roof is warmer in winter it should decrease the daytime winter warming energy required. I wonder if this has been measured or modeled in the past to determine what the benefit if any is. (Most testing - such as at the FSEC - seem to concentrate on summer performance of roofing systems.) 'Though if you design using passive solar principles perhaps you don't need this benefit.