How do I add geom_errorbars() using a separate dataset to an existing geom_violin() dataset in ggplot2?

Question:

How do you properly stack a geom_errorbar() that uses a different (but related!) dataset on top of geom_violin()?

Setup

I ran an auction receiving many individual bid prices for shipping lanes. Afterwards, I selected 2 or 3 bids to create a weighted average award price. Additionally, for each lane, I received an external benchmarking service's low and high market estimate for the lane. In a single plot, I want to show:

• geom_violin() of the distribution of received bid prices.
• geom_jitter() of the individual bid prices.
• geom_point() of the single, weighted average/blended award.
• geom_errorbar() (or related geom -- preferably shaded area using geom_rect()) that shows the market estimate between low and high for that item.

To minimize repetition, I am using 2 dataframes, one with the numerous bid prices for each item, and one that contains just the blended award and market estimates:

df_detail <- data.frame(item = c("Item_1", "Item_1", "Item_1", "Item_1", "Item_1",
                                 "Item_2", "Item_2", "Item_2", "Item_2", "Item_2", 
                                 "Item_3", "Item_3", "Item_3", "Item_3", "Item_3"), 
                        bid_price = c(6678, 3236, 4674, 5487, 5490, 
                                      5800, 6678, 3827, 4674, 5487, 
                                      2266, 1195, 1225, 3285, 1350))

df_summary <- data.frame(item = c("Item_1", "Item_2", "Item_3"), 
                         award = c(4278, 5844.32, 1244.31), 
                         mkt_low = c(4603, 4619, 833), 
                         mkt_high = c(5661, 5681, 1176))

Work So Far...

I know from other experience and reading that you can put the main dataset in the call to ggplot() and add geom_*s(), plus add separate datasets inside their own geom_*() calls. I have no trouble with getting everything except the geom_errorbar() in this call:

library(ggplot2)

ggplot(data = df_detail, # Primary dataset using all the bid prices
       aes(x=item,
           y = bid_price)) + 
  geom_violin(fill = "gray90") +
  geom_jitter(color = "darkorchid2", 
              size = 0.75, 
              alpha = 1, 
              width = 0.15) +
  geom_point(data = df_summary, # Add summary dataset with single award point.
             aes(x = item, 
                 y = award), 
             color = "red") +
  theme_bw() +
  coord_cartesian(ylim = c(0,7000))

...and I can separately create the geom_errorbars() in their own plot containing the market lows and highs for each item:

ggplot(data = df_summary,
       mapping = aes(x = item,
                     ymin = mkt_low,
                     ymax = mkt_high)) +
geom_errorbar(width = 0.5) +
theme_bw() +
coord_cartesian(ylim = c(0,7000))

The Problem

I can't put them together in the same plot:

ggplot(data = df_detail, 
       aes(x=item,
           y = bid_price)) + 
  geom_violin(fill = "gray90") +
  geom_jitter(color = "darkorchid2", 
              size = 0.75, 
              alpha = 1, 
              width = 0.15) +
  geom_point(data = df_summary, 
             aes(x = item, 
                 y = award), 
             color = "red") +
  geom_errorbar(data = df_summary,  # I am adding this geom_* call!!!!
                mapping = aes(x = item,
                              ymin = mkt_low,
                              ymax = mkt_high),
                width = 0.5) +
  theme_bw() +
  coord_cartesian(ylim = c(0,7000))

The above errors for me as seen below in the trace. However, I'm confused because I didn't have a problem with the geom_point() call which worked earlier. I don't understand the geom_errorbar() looking for the object bid_price which is part of df_detail. If I run rlang::last_trace(), my skills are not good enough to decipher this usefully:

I feel I have some major misconception here, and I appreciate any help offered. Thank you in advance.

Because you set data =df_detail and aes(y = bid_price) in the ggplot() call on line 1, they will be inherited as the default values for all subsequent geom_*.
This is true for the aes(y = ) parameter in your geom_errorbar(), regardless of the custom data and aes(x) parameter provided to the error bar.

To get what you want, either set the y parameter in each geom rather than in the ggplot(aes()) call, or simply set the y parameter of the aes() mapping in the geom_errorbar to NULL.

ggplot(data = df_detail,
       aes(x=item,
           y = bid_price)) + 
  geom_violin(fill = "gray90") +
  geom_jitter(color = "darkorchid2", 
              size = 0.75, 
              alpha = 1, 
              width = 0.15) +
  geom_point(data = df_summary, 
             aes(x = item, 
                 y = award), 
             color = "red") +
  geom_errorbar(data = df_summary,  
                mapping = aes(x = item,
                              y = NULL, # Here!
                              ymin = mkt_low,
                              ymax = mkt_high),
                width = 0.5) +
  theme_bw() +
  coord_cartesian(ylim = c(0,7000))