The fact that some people receive badly targeted ads isn't a prove for the surveillance system not working.

The question regarding ads is how well targeted campaigns are being paid for. But the ad space is relatively harmless.

It becomes relevant when not talking about ads, but surpression of citizens by state actors. If the data is being used to identified potential targets for some measures. These measures can involve drawing election districts, deciding where to build the sewage plant or library or where search for people to deport and how to influence them in political campaigns. Some margin of error has varying impact there, but it can help a lot to reduce manual selection by a lot.