An FDR of 0.001 is ridiculously stringent. It means you allow no more than 1:1000 errors. At FDR 0.1 you allow 10% of your results to be false positive - even 25% is still considered OK (meaning about a quarter of the enrichment you find are likely to be false - but the majority is still real).

ER

I do not know exactly the details of the results, but did you verify if the DEG have a difference in gene expression of at least two fold? Sometimes I have found lists of DEG with low FDR but tiny differences in gene expression. Sometimes a very small difference in gene expression cannot be biologically relevant and the list of genes does not show any Gene Ontology enrichment.

Good luck!