R dplyr: min max function not working in mutate - r

Question

I have an issue with dplyr I cannot resolve. Also I do not have a full workable example, since the problem only occurs with the full set of data (that I cannot share with you).
I do the following:
t %>% group_by(id, add=TRUE) %>%
summarise(minbplevel = min(ref, na.rm=T)
,maxbplevel = max(ref, na.rm=T)
) %>% filter(id %in% c(caseA,caseB))
Which results in
id minbplevel maxbplevel
(dbl) (dbl) (dbl)
1 B 33.0 73.0
2 A 39.4 80.4
But when I do
t %>% group_by(id, add=TRUE) %>%
mutate(minbplevel = min(ref, na.rm=T)
,maxbplevel = max(ref, na.rm=T)
) %>% filter(id %in% c(caseA,caseB))
It results in:
id Level refparmax refparmin ref meanbptest minbplevel maxbplevel
(dbl) (chr) (int) (int) (dbl) (dbl) (dbl) (dbl)
1 B 0SD 69 68 49.0 52.00000 33 73
2 B min1SD 69 68 41.0 52.00000 33 73
3 B min2SD 69 68 33.0 52.00000 33 73
4 B plus1SD 69 68 59.0 52.00000 33 73
5 B plus2SD 69 68 73.0 52.00000 33 73
6 A 0SD 100 95 56.4 35.33333 NA NA
7 A min1SD 100 95 47.4 35.33333 NA NA
8 A min2SD 100 95 39.4 35.33333 NA NA
9 A plus1SD 100 95 67.4 35.33333 NA NA
10 A plus2SD 100 95 80.4 35.33333 NA NA
Why the NA's in case A are produced, I have no clue. It seems that each time I try it on a subset of the data, the second case with data is the problem, but that is just a hunch.
It is only one case of the 18850 that gives this issue, but there is nothing identifiable that makes the problem case different than the rest.
Please advice what I can try to do to solve this?
I can think of workarounds, creating the summarized data and then merging the result with the original data. But I thought that dplyr would allow me to do this in one step.
I tried removing or adding the add = TRUE option. That does not make any difference.
Maybe I am using this in the wrong way.
Based on comment I tried:
subset(with(t,aggregate(ref~id, t, FUN= min, na.rm=TRUE, na.action= na.pass)),id %in% c(caseA,caseB))
Which results in
id ref
4 B 33.0
5 A 39.4
I have to mask some parts of the data.
dput(head(subset(t,id %in% c(caseA,caseB)) , 12))
gives:
Again I replaced the actual id's with variables caseB and caseA. Also this is not the full dataset in which the problem occurs.
structure(list(id = c(caseB, caseB, caseB, caseB, caseB,
caseA, caseA, caseA, caseA, caseA), Level = c("0SD", "min1SD",
"min2SD", "plus1SD", "plus2SD", "0SD", "min1SD", "min2SD", "plus1SD",
"plus2SD"), refparmax = c(69L, 69L, 69L, 69L, 69L, 100L, 100L,
100L, 100L, 100L), refparmin = c(68L, 68L, 68L, 68L, 68L, 95L,
95L, 95L, 95L, 95L), ref = c(49, 41, 33, 59, 73, 56.4, 47.4,
39.4, 67.4, 80.4), meanbptest = c(52, 52, 52, 52, 52, 35.3333333333333,
35.3333333333333, 35.3333333333333, 35.3333333333333, 35.3333333333333
)), .Names = c("id", "Level", "refparmax", "refparmin", "ref",
"meanbptest"), class = c("grouped_df", "tbl_df", "tbl", "data.frame"
), row.names = c(NA, -10L), vars = list(id), drop = TRUE, indices = list(
0:4, 5:9), group_sizes = c(5L, 5L), biggest_group_size = 5L, labels = structure(list(
id = c(caseB, caseA)), class = "data.frame", row.names = c(NA,
-2L), vars = list(id), drop = TRUE, .Names = "id"))

Answer 1

If I replace all NA's in the ref column with zeros the mutate step is working fine. As aosmith suggested, it has probably something to do with the mutate and NA issue that is fixed in the developement version of dplyr.
I cannot test this suggestion due to workstation restrictions though. So I will work around the issue, with the NA replacement step, and process the zero values after the summary steps.