Real readout is often asymmetric between prepared states; three rows move from a symmetric baseline to a coin-flip one-state error rate. Exact arithmetic here means exact results for the stated model inputs; measured inputs still carry uncertainty and significant-figure limits.
highlighted = computed this step
Earlier scans always moved both rows together
Chapter six's diagonal-strength scan holds the two prepared-state correct rates equal by construction and moves them together. This chapter's baseline row starts equal too: both prepared states read correctly with probability 9/10.
Azz=109=Aoo=101
This scan holds one row fixed and degrades only the other
The next two rows keep the prepared-zero correct rate fixed while only the prepared-one correct rate drops -- no earlier scan in this book holds one row fixed while moving the other alone. This is the same shape a decay-during-readout effect would leave: the excited state is harder to read correctly than the ground state.
Azz109109109Aoo10110321
The strongest row reads prepared-one as a coin flip
In the last row, prepared one reads correctly only 1/2 of the time, while prepared zero is unchanged. The two prepared states no longer share one calibration number.