SQLite does not have very many execution methods or a very intelligent optimizer, and every database has different access methods available to it to some extent. For example, Postgres lacks skip scan, which Oracle has. MySQL lacks bitmap index scans to combine indexes, which Postgres and Oracle have. Postgres (until the recently released 9.2) lacked index-only scan, which MySQL and Oracle (and DB2, and Informix, and SQL server...that was a feature long in coming) have. MySQL of many versions lacked hash joins (which blows my mind), it may have it now (depends how you count the forks).

So, all in all...somewhat expected. I have a number of queries that simply cannot run in their most naive rendering on Postgres due to lack of skip scan.