Short Answer: The memory does vary within a card model and it makes things very difficult to compare.
Long Answer: The difficulty is that there are so many variables when comparing one card to another - hashrate (depending on memory type and mining algorithm), power usage - add to that the ever changing prices and 'out-of-stock' issues. The charts on the site are an attempt to make all of that research easier, and I'm working on adding more features that will make it easier.
It's quite easy to compare different Pascal cards cause they're all pretty much the same. Nvidia nerfed serious overclocking with this generation so it doesn't really matter which brand you buy all reach ~ similar clock/voltage levels provided the cooling is good enough (and it usually is since the cards are not very power hungry). The one thing that matters the most is the memory brand. Especially for 1060s and Dagger, where the difference is up to 20-25% (Hynix = up to 20 MH/s, Samsung = up to 25 MH/s). And that's one thing that you can't predict with any articles, charts and links to the stores. People tried before. Spent a lot of time contacting manufacturers, sharing batch numbers, trying to figure out a way to predict the memory type. Same thing happened with Polaris cards before, where everyone was also hunting for Samsung chips. No luck there, it's just mostly random. There are a few exceptions, with 1060s there are special "mining edition" cards, and some of them are guaranteed to have Samsung chips installed. But with regular 1060s it's a lottery.