Web detection returns an extremely low number of matching pages and images.
Normally the results are identical, or at least very similar, to a manual Google image search but that is no longer the case.
An example, the image attached, only get 14 pagesWithMatchingImages back but it should be similar/identical to a manual Google image search that returns around 154 image sources for the same image.
Steps to reproduce: Call the Image Annotate API with the web detection feature (WEB_DETECTION / 10) and the max number of results (maxResults) set to a high amount e.g. 10000.
