In an attempt to improve retrieval systems for the patent domain, significant efforts are invested to assist researchers in formulating better queries, preferably via automated query generation. Current work on query generation in patent retrieval is mostly based on statistical measures without considering whether these terms are the best choice. To learn from actual queries being posed by experts, we analyze query logs from USPTO patent examiners. Results show that US examiners pick the majority of query terms from the claim section, a large fraction of which, in turn, coincide with the subject feature terms which determine the extent of the protection of the patent right. Considering the lessons learned from evaluating existing search logs will help in improving (semi-) automated query generation.