Positive TPM in genes with 0 counts from StringTie?

Question

[fr]

Hi,

I'm confused by the StringTie calculation of TPM (docs) for output TPM. They point to this paper for the definition of TPM (but I can't find much about it), and to this tutorial for explaining what TPM is. However I find that some genes with 0 counts show a positive very low TPM (0.15). Can someone please explain how this can be so given that the following is the calculation of TPM:

TPM = ( counts * sum_counts_in_sample) / ( gene_length_kb * 1e6)

Thanks in advance