The hole in the ground below a magma workshop can be anywhere you like, but I really prefer to have it below the impassable tile of the workshop. So yes, the weaponsmith would need the magma below the centre row. You can of course just have a 1 tile wide magma plumbing and just put little dents into it where you will probably channel out the holes. This give a vaguely comb-like appearance and can look quite neat.
It depends on the workshops I want to put there (the location for "
the tile" in magma forges being [centre-row, left and/or right] access point and all the others being [top-row, centre]) but I will generally run my single-width magmaduct(s) so that there are no 'dents' or untoward branchings at all (beyond possibly a distribution spot right next to the magma source). This also improves the dedication of the flow, given I use tapping techniques which give a relatively low-speed flooding speed, and spreading into multiple new tiles slows this down.
It's slightly less efficient if/when I want to make corners (e.g. to snake around to service a 3x3 array of magma workshops). I could either make a diagonal gap to the bend, causing its own slowing issues, or suffer it being a proper L-junction with "two spots to fill", also (when you come down to it) that latter would mean one more tile to have to completely fill before 7/7 stable (or at least 4/7 usable) the whole way through.
So these days I actually tend to 'radiate' new magmaducts off of a magma pipe, or through a handy rock mass within the magma sea level, so that they're all straight. Instead of a single 3x3 workshop array all full of workshops, I might have three 3x3 magma workshop 'areas', only the central set of trios being the workshops, the adjacent rooms being part of the supply and produce piles (as well as any Z+/-N stores I can establish, ignoring of course where the magmaduct cuts through the Z-1 plan).
This also means that I can either extend existing radiating tunnels (if the geology allows) or 'star' off some more from the pipe (on this or another level, as existing excavation/geology layout allows). But I must admit that really thorough tapping plans have never yet been required so this strategy has never yet been tested to destruction, but it appears to be generally viable up to a huge degree of expansion. (A magma pipe is essentially revealed in its usable entirety, but even when it comes to the magma sea, adding in new tapping-points grants further view of the environment down there and lets you plan more tappings further away, which, in turn... You'll only really have problems if you have essentially explored the entire magma sea, and if you haven't got a productive enough magma industry set up by then, you're obviously being a busy little dwarf and have certain Plans well beyond simply pouring out metalcrafts.
)
Hopefully I've used an understandable form of terminology and not forgotten about some other limitation that I've not yet had direct experience of.