strictly speaking, the transcript which has read count value 50 should not be 0 FPKM.

You should search and learn the formula about changing read count to FPKM, it very easy .

I suspect either a rounding error (but you have others with less counts that show FPKM) or maybe the annotation is wrong or wrongly formatted. If possible let the tool (which one are you using?) also output the length of the gene/transcript. Maybe one exon spans a whole chromosome or something similar.

The question becomes whether the transcript with a read count of 50 overlaps with other transcripts to which those reads end up being assigned. If that's the case, then yes, a transcript can have overlapping alignments but an FPKM of 0.