Optimizing a Full-Text Search Engine

"Normal database engines (MySQL, MongoDB, Postgresql, etc) did not provide the level of performance I was looking for." ... "After indexing just 358 of the 1395 files available the index was already a whopping 607MB"

I don't understand why those systems couldn't handle fewer than 2000 text documents. Or Sqlite3 + FTS3 or FTS4 for that matter.

Nor why actual full-text search engines like Lucene/Solr weren't considered. The project page suggests that the point is simplicity, but I thought Solr was simple even to get started with.