[BioC] problem with GO terms

[Ina Hoeschele]

Hi,
  I have done a simple analysis associating GO terms with a gene list using GOstats. Then when I try to retrieve all genes belonging to a significant GO category I get zero genes ! I use this code:
library(biomaRt)
mart <- useMart("ensembl", dataset="hsapiens_gene_ensembl")
temp <- getBM(attributes="entrezgene", filters="go", values=GOID[g], mart=mart)

length(temp$entrezgene) is zero!! 

GOID[g=1] = "GO:0050864", so as long as this is a valid GO ID (as returned from GOstats), length(temp$entrezgene) should not be zero!?  

This happens for multiple of my top 105 GO (BP, CC, MF) categories.

Thanks for any hint ...

Ina