I haven't found a way to directly circumvent that. My approach was to create a deduplication key in my code using the name and address, then filter out records at the end for only the unique key. It creates some waste, undoubtedly, but the brute force approach to search micro areas then clean on the back end has been the only way I've found to get the most accurate results. In high density areas especially it's even more important to have small search areas because the results are heavily influenced by the search point.

I haven't compared to foursquare results -- my two cents is that I'd probably have a bit more trust in Google Maps to have more up-to-date, comprehensive location information, but for larger search areas the foursquare could be a more efficient option.