Following from Ed’s comment about the occurrence of rhoticity in English dialects as reported by Ellis, Joseph Wright, and the Survey of English Dialects, it seems to me that Ellis was having great trouble identifying exactly what sounds he (and his principal co-workers, Thomas Hallam and Prince Louis Lucien Bonaparte) was hearing when the orthography contained ‘r’. I think we can assume that the incidence of rhoticity was falling, and perhaps quite rapidly, throughout the country during the nineteenth century, which wouldn’t have helped. Also the sound itself was ceasing to be a ‘proper’ trill or tap, and becoming more like the present-day approximant.
He recognised what we call “intrusive r” in many parts of the country, and if we had continued to use the phrase he had for it – “euphonic r”, perhaps we might have spared ourselves as linguists a lot of grief from the general public who have picked up on “intrusive r” and beaten us on the head with it for allowing what they hear as a blatant solecism to be perpetuated.
I’d also like to acknowledge here Jack Windsor Lewis’ comment about Joseph Wright’s English Dialect Grammar, and John Wells’ Accents of English. I agree that Ellis’s presentation could have been better, but as the pioneering effort it was, I think he did a great job in covering an immense amount of ground.