It happens in a similar way to anything: you see what works, and create sense with symbols and consistent repetition as in any semantic field. Why is one burger better than another, or one steak? I can't say we have the science yet for exact measurement, but it may come. Sometimes it's obvious: cheap industrial plastic supermarket cheddar really is full of less decent chemicals for your senses than pricier artisan ones like Humboldt Fog. That can be demonstrated, and I suspect in the future more will be demonstrable with wine. So far you can realise why that cheap Zin blend from Lodi doesn't really do it for you with the most obvious of measurements: yield! They get even 10 tons per acre there rather than 1-3 tons in a high-quality vineyard. Fewer nutrients and flavours to go around, and more water, sugar and untasty components have to flesh it out (poor overburdened vines, eh!) that's just the first variable: I know some more, but there are good soil scientists who can tell you a lot about all of the other variables too. Feel free to ping me for pointers, as I don't have time to write about it now.
I frequently prefer cheaper wines to more expensive ones, and agree price-quality correlation is not good, depending on what your wine source is, but at the same time you're going to have to find good data to demonstrate no correlation: for price is one (flawed but still input-data-rich) indicator!
We do fool ourselves with patterns, but also we do not. AFAIK there is not enough research to rove it either way, but qualitative methods can point us to better theories to work with until we have good enough constructs for quantitative analysis for our minds' perceptions.