If you enable office-365-connector and add a ${BUILD_FAILURE_ANALYZER} token in definitions (which would result in nice failure reason on Microsoft Teams) you get actually no failure reasons at all.

Without usage of ${BUILD_FAILURE_ANALYZER} token all failure reasons are read correctly.

I've did a little debugging of the plugin and I've found that it's due to:
1) Office365Connector sends notification on start/fail.
a) Job starts we get a request for ${BUILD_FAILURE_ANALYZER}.
2) Build-failure-analyzer recognizes that it does not have failure reasons so it tries to parse log and marks job as analyzed and as job just started there's no failures yet.
3) Job finishes so we get once more call to analyze but as it was already done it's skipped.