This is related to an expected difference between the HUMAnN v3 and v4 lineages. V3 calls MetaPhlAn in the default way and uses the resulting last column of the output, which contains relative abundance information, for species selection. V4 calls MetaPhlAn with the additional read stats option and then uses the coverage column for species selection. V3 would not be expecting MetaPhlAn output with read stats included.
That said, it’s not clear to me from your message if what you’re seeing is an error vs. the expected structure/behavior of the two lineages of code?