How to get groupwise error bars in correct place on stacked bar/dynamite plot in ggplot?

Question

I have a dataset that I want to plot the mean and confidence limits of in a dynamite plot (I am aware of the drawbacks of dynamite plots but I work in a corporate environment where simple summary statistics are often more useful than representing the distribution.) I have calculated the summary statistics prior to plotting as such:

library(gmodels)

lci <- function(data) {
  as.numeric(ci(data)[2])
}

uci <- function(data) {
  as.numeric(ci(data)[3])
}

ID <- seq(1:140)
avg_outcome <- c(rnorm(100, 100, 20), rnorm(40, 60, 30))
avg_outcome[avg_outcome<0] <- 0
group <- c( rep("Red", 100), rep("Blue", 40) )
area <- c(rep("North", 50),rep("South", 10), rep("West", 15), rep("East", 25), rep("North", 15), rep("South", 5), rep("East", 20))

reprex <- data.frame(ID, avg_outcome, group, area)

reprex_avg <- reprex %>%
  group_by(group, area) %>%
  summarise_if(is.numeric, funs(mean, lci, uci))

But I can't work out how to get the error bars to successfully stack in the way the bars are:

ggplot(reprex_avg, aes(x=area, fill=group, y=avg_outcome_mean, ymin=avg_outcome_lci, ymax=avg_outcome_uci)) +
  geom_col(color="black") +
  geom_errorbar()

gives:

I tried explicitly setting the position e.g.

ggplot(reprex_avg, aes(x=area, fill=group, y=avg_outcome_mean, ymin=avg_outcome_lci, ymax=avg_outcome_uci)) +
  geom_col(color="black", position=position_stack(0.5)) +
  #geom_errorbar() +
  #geom_errorbar(aes(group=group)) + # this does nothing
  #geom_errorbar(aes(group=group, linetype=group)) + # different linetypes but still not stacked
  geom_errorbar(aes(group=group, linetype=group), position=position_stack(0.5)) +
  geom_hline(yintercept=0) +
  theme_bw()

which gives:

along with the warning message:

Warning message:
Stacking not well defined when not anchored on the axis

Which I don't really understand as I have set aesthetics for both axes.

I know I could get around this by using a dodged dynamite plot or by faceting by group but in this case I would much rather stack, because there are a lot of different areas in my real dataset so dodging would look confusing and also because the smaller group is only present for a small subset of them, so faceting would result in a lot of unused space.

Does anyone know how I can stack the error bars successfully?