KM with CI ribbon does not show correctly with plotnine

Using default code

from lifelines.datasets import load_dd
data = load_dd()
kmf = KaplanMeierFitter()
kmf.fit(data["duration"], event_observed=data["observed"])
kmf.plot_survival_function()
plt.xlim(0, 20)
plt.show()

produces KM curve with CI ribbon as expected, attached.
However, combined with the following this plotnine code

km_df = kmf.survival_function_.reset_index()
km_df.columns = ["Time", "Survival"]

ci_df = kmf.confidence_interval_.reset_index()
ci_df.columns = ["Time", "Lower", "Upper"]
km_merged = km_df.merge(ci_df, on="Time")

(
p9.ggplot(km_merged, p9.aes(x="Time", y="Survival")) +
p9.geom_step(size=1.2, color="blue") +
p9.geom_ribbon(p9.aes(ymin="Lower", ymax="Upper"), alpha=0.25, fill="blue")
)

the CI ribbon is not computed/aligned correctly with the survival function, see attached.
Any idea what is going wrong?

Reproduce the example using the merged survival and confidence-interval data, then inspect the behavior of p9.geom_step and p9.geom_ribbon together. Compare the rendered ribbon with the lifelines plot and determine whether the issue is in the supplied data alignment or plotnine rendering; done means the confidence interval follows the survival curve correctly.