Generally RFID operates on near-field interactions. This is language that make something simple sound much fancier then it is. When you are in the near field of an antenna any coupling is due to capacitive and inductive coupling.
Most antennas for such a system are actually just large loops of wire and the tag often has a similar construction. The communication is exactly like coupling an AC wave through a transformer. So the different frequency still works through the transformer. The only issue you will have is if the frequency is too high to couple efficiently.
I'm going to examine the maths of transmitting power from a point source (the reader antenna) and assume the passive tag has an antenna that can pick-up, rectify and re-use a fraction of this power. This doesn't consider near-field effects/benefits (these may save the day but my gut feeling suggests they won't).
The reader antenna is pumping out +20dBm - that's a power of 100mW and assuming that is transmitted in all directions and your tag's antenna has an effective aperture\$^1\$ of 50 sq cm, the question is what amount of power can the tag "liberate" from the "ether"?
At 1 foot (30.4cm) distant, the 100mW output is shared over every sq cm of a sphere of radius 30.4cm. The spherical surface area is \$4\pi r^2\$ = 11,613 sq cm.
If the antenna has an effective aperture of 50 sq cm it will receive 100mW x \$\dfrac{50}{11613}\$ = 0.43mW.
It doesn't sound much but, this power can be stored (and accumulated) for about a second or so and build up enough energy in the tag's circuits enabling it to vomit a short response. That's all that is needed for this to work.
Let's make life easier - lets say the reader antenna (transmitting power for the tag) has a gain of 10dB over an isotropic antenna (an isotropic antenna produces a spherical field as in my earlier noodlings).
Now, we can say that the power received is 10 times larger at 4.3mW and this is beginning to sound reasonable but, the transmit field is much more restricted in shape but hopefully this isn't a problem.
As a side note 4.3mW across a 50 ohm load equates to an RMS voltage of nearly half a volt RMS and of course energy harvesting technology could easily reap enough energy to make a short but useful tag transmission.
What about getting it to work at 10 foot (304cm)? The effective power is now 1W because we are using an antenna with 10dBi gain. The area of the sphere is now 1.16 million sq cm and the fractional power hitting the tag antenna (50 sq cm effective aperture) is a measly 43 micro watts.
Will this work? Could it work? If the 43 \$\mu watts\$ were received across a 50 ohm resistor the voltage would be about 46mV RMS - I don't think any energy harvesting circuit will liberate anything from this.
Conclusion -
A passive tag won't work at ten foot - it would need to be powered by a small button-cell and it's MCU would need to wake up every 1 or 2 seconds to see if there was something potentially being received.
\$^1\$ The effective area of a tag's antenna is just an estimate on my part based on a little experience and the frequency the system it is operating on. Theoretically an isotropic antenna operating at 900MHz will have an aperture of \$\dfrac{\lambda^2}{4\pi}\$ which is \$\dfrac{0.1111^2}{4\pi}\$ which equals 8.84 x \$10^{-3}\$ sq metres or 88.4 sq cm. I've assumed the actual antenna's aperture is 50cm. This is somewhat less than an isotropic antenna despite the liklihood that the tag antenna will likely have gain/directionality BUT it will also have a poor efficiency because of it's tiny size and construction. A "real" quarter wave dipole antenna will be 8cm long which obviously is unattainable in a small tag.
Best Answer
...How is it possible?
Easy, it's not that tag which detemines the frequency but the reader.
The tag works between 840 - 960 MHz. Also that tag is just some antennas and a chip. The tag has no battery, it gets its power from the reader via its antennas.
The tag also does not have a way of generating a precise frequency. For that it would need a crystal which would make the tag more expensive. And there's no need for that. The combination chip + antennas is just made such that it can work between 840 - 960 MHz.
Now the reader is more complex, it also needs antennas to communicate with the tag. It needs a power source like a battery, adapter or USB connection. It will also have a crystal for generating a precise clock. This allows you to set it to a certain frequency.
As long as that frequency is within the suitable range for the tag, it can be read.
When the tag receives the signal from the reader, that signal is used to power the tag. To transfer the data some clock is needed, it is possible that the tag divides down the frequency it receives from the reader to a much lower frequency and it might use that as a clock to transfer the data. It is also possible that the tag uses its own (less accurate) internal clock and that the reader simply derives the data clock from that signal.