Because there are many benefits to using GraphQL that aren't supported if you explicitly tie the schema to the data source - one of them being the relative ease of supporting out-of-date clients. Another being the ease of supporting multiple data sources in the same schema, like Elasticsearch, Redis, etc (though I guess you can have PostgreSQL talk to those data sources instead).

My use of "real project" was perhaps a little unfair. What I mean is that as soon as you go beyond the simplest requirements of a GraphQL server, a solution like this is going to present more problems than it solves.