R ggplot - ggrepel label positions

Been trying to make a pie chart with ggrepel labels. The chart looks fine itself but the label positions are way off from what they should be.

lightingdf <- structure(
  list(
    Light = c(
      "Daylight",
      "Dark",
      "Dark, artificial",
      "Dusk",
      "Dawn",
      "Dusk, artificial",
      "Daylight, artificial",
      "Dawn, artificial"
    ),
    count = c(1291L, 209L, 186L, 71L, 29L, 18L, 14L, 13L),
    percentage = c(
      70.5079191698525,
      11.4145275805571,
      10.1583833970508,
      3.87766247951939,
      1.58383397050792,
      0.983069361004915,
      0.764609503003823,
      0.70999453850355
    )
  ),
  row.names = c(NA, -8L),
  class = c("tbl_df", "tbl", "data.frame")
)
lightingdf <- lightingdf %>% mutate(Light = str_replace_all(Light, " ",""))

light_levels <- c("Daylight","Daylight,artificial",
                  "Dark","Dark,artificial",
                  "Dusk,artificial","Dusk",
                  "Dawn,artificial","Dawn")

lightingdf <- lightingdf %>% mutate(Light = factor(Light,levels = light_levels))

baseplot <- lightingdf %>% ggplot(aes(x = "", y = percentage, fill = Light)) + 
  geom_bar(stat = 'identity', color = 'black',width = .5) + coord_polar("y", start = pi/2)

color_palette <- c(
  "Daylight" = "#27B7CE",              
  "Daylight,artificial" = "#93DBE7",  
  "Dawn" = "#FB8048",
  "Dawn,artificial" = "#FCA67F",
  "Dusk" = "#FFAA00",
  "Dusk,artificial" = "#B37700",
  "Dark,artificial" = "#47B972",
  "Dark" = "#FB6476"
)

themedChart <- baseplot + 
           labs(caption = "Collisions by light type - 2020 to 2024") +
           theme_void() + 
           scale_fill_manual(values = color_palette) +
           theme(plot.caption = element_text(hjust=0.5,vjust=10,size=14,face="bold"),
                 plot.margin = unit(c(1.2,1,1.2,1),"cm"),
                 panel.background = element_rect(fill = "white"),
                 legend.position = "none")


positionDF <- lightingdf %>% 
  mutate(
    csum = rev(cumsum(rev(percentage))),    
    pos = percentage/2 + lead(csum, 1),     
    pos = if_else(is.na(pos), percentage/2, pos) 
  ) 

themedChart + 
geom_text_repel(data = positionDF,
                aes(y = pos,
                    label = paste(Light,round(percentage,1),"%",sep=" ")),
                size = 3,
                color = "black",
                fontface="bold.italic",
                nudge_x = 0.75,
                show.legend = FALSE)

I'm getting this as output

But the 'leader lines' from ggrepel don't match up with my factor. Ex: The line for "Dawn" is pointing to the section for "Dusk", and "Daylight,artificial" is on the right of the plot while the wedge that corresponds to it is the light blue one.

The output i'd like to get would look something like the following:

I'm fine with some of the smaller labels (<1%) not being placed, but I want this order.

I tried rearranging light by factor and recalculating pos, but got the same output. I tried manually adjust the pos positions, but I doubt that's the 'right' approach to take.

The issue is your computation of the positions for the labels, which can be easily seen when plotting in cartesian coordinates.

After fixing the computation of the positions for the labels your code works fine.

library(ggplot2)
library(ggrepel)
library(dplyr, warn = FALSE)

positionDF <- lightingdf |>
  arrange(desc(Light)) |>
  mutate(
    csum = cumsum(percentage),
    pos = (csum + lag(csum, default = 0)) / 2
  )

baseplot <- lightingdf |>
  ggplot(aes(x = "", y = percentage, fill = Light)) +
  geom_col(color = "black", width = .5) +
  geom_text_repel(
    data = positionDF,
    aes(
      x = 1.2,
      y = pos,
      label = paste(
        Light,
        scales::number(percentage, accuracy = .1, suffix = " %")
      )
    ),
    size = 3,
    color = "black",
    fontface = "bold.italic",
    nudge_x = 0.3,
    show.legend = FALSE
  )

themedChart <- baseplot +
  scale_fill_manual(values = color_palette) +
  labs(caption = "Collisions by light type - 2020 to 2024") +
  theme_void() +
  theme(
    plot.caption = element_text(
      hjust = 0.5, vjust = 10,
      size = 14, face = "bold"
    ),
    plot.margin = unit(c(1.2, 1, 1.2, 1), "cm"),
    panel.background = element_rect(fill = "white"),
    legend.position = "none"
  )

themedChart +
  coord_polar("y", start = pi / 2)