It might be best to use the max that is at the maximum observed time or less. Like "breaks = seq(0, 24, by = 6)" for the example above to make it look pretty. To force it in assumes the same cumulative incidences of Drug A and B for month 23.9 to be the same as month 25.00 if you look at summary(survfit2(Surv(ttdeath, death_cr) ~ trt, tidycmprsk::trial)). It's NA because the number at risk is 0 at month 25 and dividing by 0 is undefined.