As of this PR : #1128, we can calculated expected limits/bands assuming non-zero signal strengths for the background hypothesis (without changing the definition of CLs).
It is not possible to do this with toys, although it should be (fairly) straightforward to implement in the code. It would require producing an additional distribution of the test-statistic, generated using toys at a user-specific signal strength, from which median and other quantiles are determined, and used in the expected upper limit calculation.
As of this PR : #1128, we can calculated expected limits/bands assuming non-zero signal strengths for the background hypothesis (without changing the definition of CLs).
It is not possible to do this with toys, although it should be (fairly) straightforward to implement in the code. It would require producing an additional distribution of the test-statistic, generated using toys at a user-specific signal strength, from which median and other quantiles are determined, and used in the expected upper limit calculation.