Yes, you can record up to six tracks, one of them being an auto accompaniment track.
However, you cannot do any “overdub” or “punch in/punch out” recording, by which I mean the ability to record over just a small portion of a track if you make a mistake.
If you do make a mistake, you have two options:
(1) You can erase that track— not the entire song, just the track with the mistake in it— and record it all over again, erasing and re-recording that track as needed until you’re finally satisfied with it.
(2) You can leave the mistake in the track. When you’re done recording all of the tracks for the song, you can use the “Save SMF” function to convert the User Song into a Standard MIDI File (SMF), load the MiDI file into a DAW, and use the DAW to correct any mistakes. Then you can save the corrected MIDI file and play it back on the keyboard.
As far as the individual tracks, you don’t need to record an auto accompaniment track, but if you do, you can either record it by itself, or record it at the same time as the first melody track. Recording it by itself has the advantage of letting you play with both hands when you record the first melody track, but it also has the possible disadvantage that it might be more difficult for you to keep the timing (or measures) straight in your head while recording just the accompaniment. Either way, the track will not record the actual accompaniment, just the section changes and chord progressions which control the accompaniment. If you use the “Save SMF” function to convert the User Song to SMF, then the actual accompaniment will be written at that time to the SMF, complete with all of your section changes and chord progressions. Since an accompaniment (style) can contain up to eight channels, that means the accompaniment track will be expanded to as many as eight channels.
Each of the five melody tracks can contain either one or two channels, but they aren’t independent, because if you do record two channels on a single track then they will be layered together— in other words, the Main Voice plus the Dual Voice— so both voices will be playing the same notes as each other, although you can set them to play in different octaves.
You cannot record the Split Voice, which at first might seem like a terrible inconvenience, but it can also be a blessing. For one thing, the Split Voice can’t be modified in all the same ways that the Main Voice and Dual Voice can (namely, changing the Attack Time, Release Time, Filter Cutoff, and Filter Resonance). And if you want to record a bass or “left-hand” track, you can still do that— you just need to play and record the track using the Main Voice, or the Main Voice plus the Dual Voice. The clear advantages of this are that you aren’t limited to using the keys which are to the left of the Split Point, as you would be if you were recording with the Split Voice; you can use both hands for playing and recording the track, instead of being forced to use just your left hand; and you can get a better variety of sounds by being able to layer two voices together if desired and modify more sound parameters.
An important thing to keep in mind is that the keyboard has a limited amount of polyphony, meaning the number of notes it can play at the same time. This means you shouldn’t go overboard when you record each track of a multitrack song, because you don’t want the total number of notes for all of the tracks to exceed the keyboard’s maximum polyphony. And if you plan to play a “lead” part on the keyboard while the User Song or SMF is being played back, then you want to make sure the recorded tracks have left enough unused notes (or tone generators) available for you to play the lead. On the other hand, if you’re going to be playing your guitar while the keyboard is playing back the recorded tracks, then you can use all of the polyphony for the recording.