I'm initially using a combination, for example, I would often start by putting the melodic note pitches on paper along with the bass note or harmony (leaving the details of the accompaniment for later). Then I'd go through that, defining where the bar lines are, then assigning the rhythmic values to the individual melodic notes. After that I would input what I have at that point into the notation software, print it out, and annotate the accompaniment into the printout, then return to the notation software, input the accompaniment, and finally put in dynamics and expressive markings. That is a fair summary of how I did the first movement.
With something highly metrically regular, like the closing march, I would do most of it in the notation software, because being able to cut and paste matching or near matching bars really does save a lot of time.
I will say that having editing software which has the ability to play isolated, specific frequency ranges on their own makes the aural dictation aspect much easier. I didn't really have to do much in the way of slowing down in order to extract the notes. There are a few places where I've notated what I meant to play, as opposed to what I actually did play.