Hmmm
I can picture what you are saying and I have to say that is a very interesting concept and one that has occupied my mind for the last hour here at work.
I haven't done any calculations but after thinking about it I decided that you would only reduce the error by half. This is because you are effectively butting together two right angled triangles with the threaded rod in each only having to turn at half the normal speed to maintain the required rate. (Gosh I wish I had a picture my brain is hurting again.
)
Imagine the case where you split the rod into smaller sections with more universal joints connecting them. The tangential error will get less and less until the case where you have an infinite number of sections of threaded rod (this is exactly what a curved threaded rod is) which will then have zero tangential error.
Phew I need a lie down!!