Slope credible intervals do not cover slope estimate

[stweb75]

I have found the slope sign probabilities (slpSgnPosPr ...) easier to interpret than the slope Credible Interval (slpCI).
Nonetheless, the slpCI is often wrong in that the slope estimate (slp) is outside of the CI.

## Example 1; trend only model
library(Rbeast)
data(Nile)
out <- beast(Nile, season="none", dump.ci=TRUE)
plot(out)
# trend CI looks OK
plot(out[[7]][13]$Y, type="l", ylim=c(700,1200))
  lines(out[[7]][15]$CI[,1], lty=3)
  lines(out[[7]][15]$CI[,2], lty=3)
# slope outside slope CI at a few indices 
plot(out[[7]][16]$slp, type="l", ylim=c(-10,10))
  lines(out[[7]][18]$slpCI[,1], lty=3)
  lines(out[[7]][18]$slpCI[,2], lty=3)
which(out[[7]][18]$slpCI[,2] > out[[7]][16]$slp)  
which(out[[7]][18]$slpCI[,1] < out[[7]][16]$slp)'

## Example 2; trend + season model
data(googletrend_beach)
out <- beast(googletrend_beach, dump.ci=TRUE)
plot(out)
# trend CI looks OK
plot(out[[7]][13]$Y, type="l", ylim=c(55,75))
  lines(out[[7]][15]$CI[,1], lty=3)
  lines(out[[7]][15]$CI[,2], lty=3)
# season CI looks OK
plot(out[[8]][13]$Y, type="l", ylim=c(-20,30))
  lines(out[[8]][15]$CI[,1], lty=3)
  lines(out[[8]][15]$CI[,2], lty=3)
# season Y within CI  
which(out[[8]][15]$CI[,2] > out[[8]][13]$Y)  
which(out[[8]][15]$CI[,1] < out[[8]][13]$Y) 
# slope frequently outside slope CI
plot(out[[7]][16]$slp, type="l", ylim=c(-10,10))
  lines(out[[7]][18]$slpCI[,1], lty=3)
  lines(out[[7]][18]$slpCI[,2], lty=3)
which(out[[7]][18]$slpCI[,2] > out[[7]][16]$slp)  
which(out[[7]][18]$slpCI[,1] < out[[7]][16]$slp)