According to the documentation, when the argument levels = "labels", then "unlabelled values become NA" which is ok.
However, when levels = "values" I would have expected to obtain a factor having all values, labelled or not.

library(haven)
set.seed(1234)
x <- labelled(sample(1:5, 20, replace = TRUE), labels = c(Good = 1, Bad = 5))

as_factor(x, levels = "values")
#  [1] <NA> <NA> 5    <NA> 1    5    <NA> <NA> <NA> <NA> <NA> <NA> 5    <NA> <NA> <NA>
# [17] 5    <NA> 5    <NA>
# Levels: 1 5

If the current functionality is alright, the documentation should mention it drops values without labels (just like the documentation for levels = "labels"), or otherwise allow as_factor() to use all values, which is what I would expect.