Hi, we had the same issue. Now our migration tool now tells us that we can go ahead (with caution). We created a new index .kibana-reindexed, with a mapping stolen from a newer .kibana index, and reindexed into it. The reindexing was done via logstash, similar to https://gist.github.com/markwalkom/8a7201e3f6ea4354ae06 ). We dropped the inconsistent .kibana, and reindexed back again. For some reason, the .kibana-reindexed also had issues. Which didn't matter, because the new .kibana doesn't.
Update: I don't recommend this. The migration to elasticsearch 2.2 worked, but the .kibana index is unusable and all objects (searches, visualisations, dashboards) are lost.