He are receiving some tweets with a language not prsent in ISO 639-1 Tables
In this page https://dev.twitter.com/rest/reference/get/search/tweets is indicated that lang par is from ISO 639-1, some we assume that the results comes from the same standard.
One example is the following tweet
Where the language is "in", which language is that?
We have around of 4k tweets from an scenario of 300k with that language.