stat_bin2d with drop=FALSE still drops some 0-count cells

[alchenist]

The drop parameter in stat_bin2d does not appear to work as advertised: docs state that "if [doc is] TRUE removes all cells with 0 counts." Given this, my expected behavior when drop = FALSE was something akin to the bottom plot. Instead, drop = FALSE results in the second plot, which does keep 0-count cells in x- or y-bins with some values but omits them in bins with no values.

Is this expected behavior? Or should bin indices that don't appear at all in the data be added back into the output of ulevels applied to bin indices in tapply_df?

library(dplyr)
library(tidyr)
library(ggplot2)

data(iris)

iris %>%
  ggplot(aes(Petal.Length, Petal.Width)) + stat_bin2d(aes(fill=..count..), drop=TRUE)

iris %>%
  ggplot(aes(Petal.Length, Petal.Width)) + stat_bin2d(aes(fill=..count..), drop=FALSE)

iris %>% 
  mutate(
    bin_x = cut(Petal.Length, breaks = 30),
    bin_y = cut(Petal.Width, breaks = 30)
  ) %>% 
  count(bin_x, bin_y) %>% 
  tidyr::complete(bin_x, bin_y, fill = list(n = 0)) %>% 
  ggplot(aes(bin_x, bin_y, fill = n)) + geom_tile()