Thanks @paulromano for looking into this.
Yes, I see agreement between the OpenMC, MCNP, and Serpent.
One small caveat is that MCNP doesn’t resolve well the peaks around 10 keV, but OpenMC and Serpent do.
I have another question about sampling the source position.
Can I use the method described above if my geometry consists of two concentric spheres and the source is in one of them only?
Is there a way to choose the cell where the particles are born?
I’ve seen this other post.
I think I need something along those lines.