This sort of physical characteristic fingerprinting has been done for years on magnetic stripe cards and EEPROM smartcards, so this is nothing new in theory, just in what physical characteristics are being measured.
In mag stripes, the magnetic remanence of the strip is different from card to card, in EEPROM, differences in the voltage levels and speed of reading of the cells are used.
The general principle is that it's no point having unbreakable crypto if the data can simply be copied to a new medium. Consider a card (of whatever type) that stores monetary value for public transport or photocopying or whatever: Put $100 on it and copy the data, not knowing which bits are what. Copy that data onto a heap of cards bought with $5 of credit on them and sell them in the grey market for $50 each and pocket the profit.
With this sort of technique, though, part of that encrypted data is a fingerprint based on the physical characteristics of the original card. The new cards will generate a fingerprint in the reader that doesn't match the original, making the copies invalid.
Sure, if you can crack the encryption, this method is useless, but that's not the point. Crypto can be pretty good and costs more than a cheap reader/writer to break to duplicate cards/RFIDs.