How large is the sample, and when were the completed deals actually agreed? June closings may reflect negotiations from an earlier market. I’d also separate serviced apartments from ordinary apartments and houses needing work; otherwise the change in property mix could create an apparent market...
I wouldn’t merge every duplicate immediately. Keep the versions side by side long enough to note the advertiser, stated update date and differing fields. One version may disappear while another remains. That gives you a practical freshness test rather than relying on the status label alone.