Zero FPKM value for transcript in Stringtie

[nachtwuenscher]

I just analyzed a dataset of 49 samples and found a zero FPKM-value in 11 of these in my favourite gene (SOX2) using texpr()-data extraction. Looking at the exon wise read counts they show nearly exact the same median value as the samples that had FPKM values > 0 (i.e. rcount value of 7800, FPKM of 41, the transcript has only one exon).
What might have had happened?
SOX2 has only one isoform and consists of only a single exon in my annotation.
Thanks for your help!