How is the interval for the fitted value computed (in the presence of outliers)?

[stweb75]

When hasOutlier=TRUE Rbeast will plot a wide gray band around the fitted Y values.
How is that interval computed? What does it mean?
I have made three attempts in the example below but can't reproduce the wide interval in the Rbeast plot.

library(Rbeast)
data(Nile)
out <- beast(Nile, season="none", dump.ci=TRUE, hasOutlier=TRUE)
plot(out)
# trend component
plot(out[[7]][13]$Y, type="l", ylim=c(700,1200))
  lines(out[[7]][15]$CI[,1], lty=3)
  lines(out[[7]][15]$CI[,2], lty=3)
# outlier component
plot(out[[9]][11]$Y, type="l", ylim=c(-200,200))
  lines(out[[9]][13]$CI[,1], lty=3)
  lines(out[[9]][13]$CI[,2], lty=3)
# no issues here
which(out[[9]][13]$CI[,2] > out[[9]][11]$slp)  
which(out[[9]][13]$CI[,1] < out[[9]][11]$slp)
# attempt 1: interval too narrow; ignores outlier uncertainty
plot(as.numeric(Nile), col="gray", ylim=c(400,1400))
  lines(out[[7]][13]$Y+out[[9]][11]$Y)
  lines(out[[7]][15]$CI[,1]+out[[9]][11]$Y, lty=3)
  lines(out[[7]][15]$CI[,2]+out[[9]][11]$Y, lty=3)
# attempt 2: interval too narrow and it's wrong to add CI limits together
plot(as.numeric(Nile), col="gray", ylim=c(400,1400))
  lines(out[[7]][13]$Y+out[[9]][11]$Y)
  lines(out[[7]][15]$CI[,1]+out[[9]][13]$CI[,1], lty=3)
  lines(out[[7]][15]$CI[,2]+out[[9]][13]$CI[,2], lty=3)
# attempt 3: interval too narrow; I'm not sure if outlierSigFactor is relevant
plot(as.numeric(Nile), col="gray", ylim=c(400,1400))
  lines(out[[7]][13]$Y+out[[9]][11]$Y)
  lines(out[[7]][13]$Y+out[[9]][11]$Y+out[[7]][14]$SD*2.5, lty=3)
  lines(out[[7]][13]$Y+out[[9]][11]$Y-out[[7]][14]$SD*2.5, lty=3)