The model is not a full fibre-tracing model, I'm afraid. You can break out one rear port to multiple front ports, but not vice versa.
If you can work within the limitations, you could treat your patch panel as duplex all the way: that is, have ports called "1-2", "3-4", "5-6" etc, where each represents a pair of LSH ports. But this won't work if you're also using single fibre paths.
Now, here's the really horrible and tedious way.
* Create a device type "lch-to-2lsh"
* Add rear port: Name LC, Type LC, Positions 2
* Add front port: Name LSH[1-2], Type LSH, link to LC:1 and LC:2
* Make an instance of this adapter, let's call it "adapt1"
* Make a connection from switchA (SFP interface) to adapt1 rear port (LC)
* Make a connection from adapt1 front port LSH:1 to fibre panel front port 1
* Make a connection from adapt1 front port LSH:2 to fibre panel front port 2
Repeat all this for the other end. This assumes your fibre panel rear ports are connected together, of course. (In my case I made a device type with 1 rear port of type MPO, linked to 24 front ports of type LSH. Then I could join the two panels with a single MPO to MPO connection)
Unfortunately, even if you do all this, you'll find that Netbox's fibre tracing does not handle the bifurcation of connections like this. When you trace the path which includes a rear-to-front path, it only traces the first leg:

This may be good enough, because at least you can *record* the connections. I might raise an issue for it anyway; but really what you want here is a *cable* with multiple connectors at one end, not a ficticious adapter.
In my opinion, the UI is still way, way too painful to do this for real. And note that in the example above, I've assumed you've connected an SFP "interface" directly to an LC "rear port". If you were being accurate, you'd also add a device of type "LX SFP", which has an SFP rear port and LC front port.
And also you have to give all these device types a rack height (e.g. 1U) and depth (full/not full), even thought you're not going to rack them. Having a rack height of "unrackable" would be an improvement there.