Confidence interval not covering the entire x axis

The confidence interval on my scatter plot doesn't cover the entire x axis, while the regression line fits the entire data set. The data that isn't covered by the c. interval doesn't have any missing values or zeros.
I tried using fullrange = TRUE , but it doesn't work.
Here is the code I used to create the plot:

plot_wet_arthropoda <- ggplot(joined_data_arthropoda, aes(x = CCL_n_t, y = total_wet_mass_arthropoda)) +
   geom_point(color = "blue", size = 3) +
  geom_smooth(method = "lm", se = TRUE, fullrange = TRUE, color = "red" ) +
  annotate("text", x = min(joined_data_arthropoda$CCL_n_t, na.rm = TRUE), 
           y = max(joined_data_arthropoda$total_wet_mass_arthropoda, na.rm = TRUE)*1.9, 
           label = annot_arth_wet, hjust = 0, size = 4) +
  labs(
    y = "Total Wet Arthropoda Weight (g)"
  ) +
  scale_y_continuous(
  breaks = c(50, 100, 200, 300, 400, 500, 600, 700, 800, 900),
limits = c(0, 900)) +
  scale_x_continuous(
    breaks = c(25, 30, 35, 40, 45, 50, 55)) +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5))

As described in https://r4ds.hadley.nz/communication.html#zooming, setting limits in scale_x_* and scale_y_* will subset the data to the plotted elements that are fully within that range. So the confidence interval part of your geom_smooth will not show portions that exceed those ranges. Below, we see the shaded area stop right when it would go below 8.

In contrast, coord_cartesian lets you specify the "viewport" of what range should be visible, without removing plotted elements that exceed that range.

Compare:

library(ggplot2)
ggplot(mtcars, aes(x = wt, y = mpg)) +
  geom_point() +
  geom_smooth(method = "lm", se = TRUE, fullrange = TRUE) +
  scale_y_continuous(limits = c(8, 35), 
                     breaks = c(8, 10, 20, 30), minor_breaks = NULL)

ggplot(mtcars, aes(x = wt, y = mpg)) +
  geom_point() +
  geom_smooth(method = "lm", se = TRUE, fullrange = TRUE) +
  coord_cartesian(ylim = c(8, 35))

Further, adding clip = "off" to coord_cartesian will show all the plotted data without clipping to the plot area. I use this often for geom_text layers where I want the text to overflow beyond the plot area without getting clipped.