I don't think you would have dialectric properties between the RFID and windshield since glass is a terrible conductor. I'm willing to bet you could even hang it on a short string on your rear view mirror and it would read just fine. I'm sure we will hear more ways to beat sticking it on as they become more prevalent. I've still not seen one yet, but I've not been looking for it either.
As I mentioned, there are several papers on the subject available on google scholar that claim otherwise.
Intuitively, I agree with you. Obviously RFID tags can work fine without being stuck to anything, given that one often finds them inserted in books and what have you these days. I presume the difficulty has something to do with the antenna being designed to be stuck to a windshield and optimized for that particular condition.
Hmm,
this patent seems to give an overview: "The antenna has a characteristic impedance defined in part by a dielectric constant of the flexible circuit substrate and in combination with a dielectric constant of the glass surface. As a result, proper impedance match is achieved only when the sticker is affixed to the glass surface"
So the range issue when it's not stuck to a windshield or something is due to an impedance mismatch in the antenna dissipating power in the antenna instead of transmitting it back to the reader.
Edited to add:
More from the patent (the numbers make sense if you look at the drawings):
Referring now to FIG. 5, an exemplary circuit schematic of the sticker transponder 10 is shown. The antenna is represented within the dotted box 60. The antenna 60 generates an alternating voltage from the signal transmitted from an RFID reader (not shown) represented by a generator 68. The impedance of the antenna 60 is represented by a capacitor 64 in parallel with a resistor 62 and an inductor 67. An impedance matching section 56 connects the antenna 60 to a transponder circuit 50, which includes a modulator 54 and a memory 52. As described above, the tuning stubs (i.e., regions 36,37, and 45, 47) provide the impedance matching section 56.
The exemplary circuit schematic further includes a capacitor 66 coupled in parallel with the impedance defined by the capacitor 64, resistor 62 and inductor 67. The capacitor 66 is a function of the dielectric constant of the glass material 24 (see FIG. 1) having a thickness within a known range. When the sticker transponder 10 is affixed to the glass 24, the capacitor 66 is in parallel with the capacitor 64, resistor 62 and inductor 67. Conversely, when the sticker transponder 10 is not affixed to the glass, such as prior to use or after removal from the glass, the capacitor 66 is uncoupled from the capacitor 64, resistor 62 and inductor 67. Accordingly, the addition of the capacitor 66 directly affects the impedance of the antenna 60. The impedance 55 matching section 56 is therefore intentionally tuned to match the relatively high impedance of the transponder circuit 50 to the relatively low impedance of the antenna 60 when the sticker transponder 10 is affixed to the glass. Conversely, an improper impedance match is formed when the capacitor 64 is uncoupled from the capacitor 64, resistor 62 and inductor 67. As a result, the antenna 60 is only effective in communicating signals with the transponder circuit 50 when the sticker transponder 10 is affixed to the glass.