Your question is typical of a popular misunderstanding of scientific data.
The statement that all snowflakes are different is meaningless unless accompanied with statistical data - if you have a data set of 1000 samples it is probability true with a fair confidence. If you could have a much larger data set of 1000 million the possibility if matching two would be higher. But for most people does it matter? I wont lose any sleep over it.
We are bombarded with these sorts of statements, most of them are sorted out by our brains and filed in the 'it wont happen to me' compartment.