You're failing to take the fall-off into account. It's true that the light is brighter in areas reached by both torches, but the there's less of each torch's light the further you get from the torch, so those overlapping areas are not necessarily the brightest spots because they're the sums of less light overall.
It is good that you've dropped the dithering for now and are focusing on the shapes of the light, that makes things easier

You can always add dithering later.
Here's a probably-not-physically-correct edit:

Most importantly, look at how much light the columns get compared to everything else. They're the closest to the fire, they get the most light. Everything else should be dimmer in comparison. (The thin highlight lines on the columns are the column ridges being highlighted, they have nothing to do with the shape of the flames.)
Notice how there's a shadow spot in the middles of the first few steps, and areas of less light. away from the middle. This makes it look like there are two light sources interacting.
This is all simpler to illustrate on the flat wall (I removed the texture for clarity). Start with the illumination created by one torch:

Note: these examples have incorrect shapes for the light, I just wanted to quickly illustrate how it works; the correct shapes would be circles with the torch at their centre. In addition, I used the available palette for the circles just to be able to show more circles, in reality the brightest colour probably wouldn't reach the wall at all.
Concentric rings of light clearly show the fall-off. You could smooth these with dithering, but I recommend saving that for a final polish step.
Then, copy this for the other torch, and add the two:

Things to notice:
- the irregular shape of the brightest part. The two slightly-weaker rings of light add up, creating a spot in the middle that's as bright as the two brightest circles of light.
- The faint light from the other torch slightly fills in the cast shadow from the column, it's not just a solid dark area.
Though you don't need to, you could extend this to more distant "slices" of the spheres of light in order to "calculate" the lighting on the stairs. I think once you understand how to do it on just a wall though, you should be able to guesstimate it without going all out. Just don't forget that the light falls off with depth too, which basically translates to the light-circles getting smaller/dimmer, and the brightest part of the circles will disappear pretty quickly.