I would disagree with the other two answers, as you are working with more than just one census record. If you put them all together, I come to the same conclusion as you: "BA" Harris in the 1870 census is an error by the census taker and that she is the same person, "Decy" in 1880, "Rhoda F." (where does Sarah come from) in the 1900 census, "Dicana" in 1910, and "Doorn" Harris in 1920. Also "Doom" Harris in the death certificate of her daughter.
The 1900 census of her father Thomas states that he had 6 children, 4 living. They are all accounted for by the 1870 and 1880 censuses. Despite the fact that her name is different in every single record ( I have someone like this in my ancestry also), there is no room for "BA Harris" and "Sarah (sic?) Frances" to be different people.