The points made above about "don't use ORM for reports" aside (a point I strongly agree with, btw), I've found Hibernate makes my life a lot simpler and rarely gets in the way or fails to do what I need it to do. But I do occasionally create a native query to deal with some edge case or whatever.
A good ORM should be able to handle a database designed for a different application, normalized to 5NF or BCNF, using composite primary keys, composite foreign keys, and other solid db design concepts, and be able to reliably generate queries to bridge the gap, and should do so where sets of objects are required in a way that does not unduly burden the database with small queries that could be consolidated.
That's a hard problem. There are some ORMs that come pretty darned close if not there. Many more, in my experience fall well short.
The second of those is what I saw at my last job. The tech-lead / architect guy was vehemently anti-ORM... because a previous group (at the same company) had rolled their own ORM in Jython and embedded it into the Java codebase, and it A. didn't perform well and B. was extremely fragile and hard to modify. Now, you can't explain to him that there are well written, productive, performant ORMs like Hibernate. He just doesn't want to hear it. sigh