It does start with the most recent and then it works backwards. I guess it stops when it reaches "relevance," whatever that means to Twitter. The results are certainly incomplete, as you've noticed.
I suspect that Twitter truncates results to limit the load on its servers, which is understandable. Their API service is free, and they can't have an unlimited number of people asking for an unlimited amount of data. Instead, they provide a subset they hope is relevant to some people. Unfortunately, it sounds like that won't give you what you need.
There are third-party providers of complete Twitter datasets, but they charge for it.
-- Tony
I suspect that Twitter truncates results to limit the load on its servers, which is understandable. Their API service is free, and they can't have an unlimited number of people asking for an unlimited amount of data. Instead, they provide a subset they hope is relevant to some people. Unfortunately, it sounds like that won't give you what you need.
There are third-party providers of complete Twitter datasets, but they charge for it.
-- Tony