I looked through some of the code for plot.acf() and if I'm reading things correctly I don't think the function makes confidence intervals for covariances.
To look at the underlying code I ran stats:::plot.acf in my Console. This is the function that plots the output of acf(), and it is this function that controls the creation and display of any confidence intervals.
Here are the lines that I thought were relevant.
First, a logical vector based on what type of acf being computed (which is FALSE if the type is covariance):
with.ci <- ci < 0 && x$type != "covariance"
Then the calculation of initial confidence interval limits based on the logical vector above. Looks like if the result is FALSE this returns c(0, 0), which is why I think you don't get CI for covariances.
clim0 <- if (with.ci)
qnorm((1 + ci)/2)/sqrt(x$n.used)
else c(0, 0)
There is a lot more code in the function so I could have missed something, but I didn't see anything else where the function built CI for covariances.
If you know how to calculate a confidence interval for a covariance you could potentially add these on to the plot via lines(), since plot.acf is built on base R plotting functions.