OK. So, in the original experiment the beam was split spin-up/spin-down, all spin-down was blocked, the remaining beam of all spin-up was then split spin-left/spin-right, all spin-right was blocked and the remaining beam of (supposedly) only spin-up/spin-left was again tested for spin-up/down and found 50%-up, 50%-down.
It's like the quantum number for spin-up/down was reset during the spin-left/right test. (reset from "up" to "?", please excuse my layman's language)
In experiment "A" after the initial spin-up/down test the beam of (supposedly) all spin-up is split spin-left/right then brought back together and the second spin-up/down test yielded the same 50-50% result. Again it's like the quantum number for spin-up/down was reset. But, I note there is a detection of spin-left/right in the experiment "A" scenario.
In the "B" experiment this detection of spin-left/right is not made and the result of the second up/down test yields 100% "up". Even though the beam in the spin-left/right apparatus is split into separate spin-left and spin-right paths (since the equipment is exactly the same, functions just the same (minus the detector)) the quantum number ("up") from the first up/down test block was not reset by the left/right test block.
The only difference between the setup of the original and "A" experiments and the setup of the "B" experiment is that the original and "A" detect the left/right split while "B" does not.
This is a "Double Slit" experiment for electron spin, yes?
So in a layman's analogy (not totally accurate but close enough for some understanding) is saying ...
"When the spin-left/right paths are detected the quantum number for spin-up/down is reset to "?" thus allowing the second up/down test to spread the results among the quantum probabilities" ...