The problem with consumption is only marginally related to the antenna in this case. The only dependency is that with a better antenna you can decrease the error ratio, reducing the total amount of sent data. But if you want to reduce significantly the power, you have to play with the power management options of the transceiver, the chapter 5 of the manual.
Probably you have already read it and optimised the application, but if you want to sacrifice range and reduce power, the most obvious solution is to lower the transmit power, as described in section 5.5 (page 18).
The cantenna reduces only the power irradiated in other directions, but not the overall power irradiated; moreover, changing the antenna, yes, can affect the power consumed by varying the antenna impedance and lowering the power transferred, but you pay it with a loss in efficiency.
IMO the best thing to do is to optimise the radiation surface of the antenna and then reduce the transmit power to the working minimum.
Note that if you decrease the power too much, it may result in a high number of lost packets, and then in a higher consumption.
I assume you have a small beacon PCB that you haven't designed yourself, so you need to work with it as-is. if you're laying out the board, you have more options...
1. Restricting the area covered by a beacon
You're onto the right track here - mounting it above the supermarket aisle, or what have you, is probably best, then you can focus the signal into a small area. You might have to reduce the range or power of the signal somehow too. It's possible that only reducing the range would be sufficient.
2. Focussing the signal
Anido is right, you can focus the signal with some sort of antenna. As you probably don't have an antenna connector on the board, you will do best with a horn type antenna, like a pringles can, that is excited by the internal antenna on the beacon. A simple open-ended waveguide without a cone flare will give an 8-10 dBi antenna, which has about a 60-degree beam (at -3 dB), or maybe 90 degrees at -10 dB.
It must of course be metal, but the type of metal doesn't matter - aluminium foil stuck to cardboard, or a tin can. The metal must be continuous, without gaps, if you need a seam, have a good 1 inch / 2 cm overlap.
Look around the net for appropriate dimensions for the horn - the diameter matters more than the depth, so a soup-tin shape is OK. You could make a pyramidal horn out of cardboard too. No need to ground the transmitter to it - it's a reflector and concentrator. (You wouldn't ground a telescope mirror...?)
3. Measuring what you have done
It is fortunate that with BLE, you have an RSSI or signal strength indicator that is quite accurate. Invent a method of recording the signal strength, maybe using your phone or another device to record the result.
Mount the new transmitter-in-can on the ground, outdoors, facing up. You don't want any reflections from the ground to interfere with the signal. Move your phone or receiver over the front of it, perhaps 1 m away, and record the results. Keep your body far away from the test - use a stick to hold the phone/receiver, so you don't affect the results too much. The exact range isn't too important, it's the shape of the pattern you're measuring; it'll be almost the same 3m away as 1m away, just 10 dB weaker.
It's also important when measuring to worry about the polarisation of the signal, and the pattern of the receiving antenna. You can't change these, but you can hunt at each location for the phone orientation with the strongest signal.
From these results you will see if you're getting the desired focussing from the can-tenna that you've invented. If it's not working, try re-orienting the transmitter in the can, or moving it closer to or further from the back wall.
So in summary: Transmitter on the ground, outside. Move phone around slowly above it, to get an idea of the shape of the radiation pattern. At each position, twist the phone around to get maximum signal. Adjust beacon position in tin can, until you like the pattern.
4. Reducing the range
You'll probably find that the signal spills over where you want it to go, even (especially) with the antenna focussing it. Time to reduce the signal power. You might be able to do this in software, on the BLE beacon, you might have to resort to some hardware modifications. Soldering a resistor in parallel with the antenna on the beacon will reduce its range, 10 ohms might be a good starting point. If you're not that brave, you could try wrapping it slowly in aluminium foil (after insulating the board, of course) to reduce the amount of radiation by creating a mismatch in the antenna. Even just placing it flat on the base of the tin, or right in one corner, might do the trick.
5. Experiment!
Apart from the can dimensions, everything else is negotiable. Experiment! Remember that you can always measure the signal strength with your receiving device, to compare before/after, and see what you've done.
Best Answer
First, to test feasibility, simply connect the inside and outside antennas together, using something such as a pair of panel-mount coaxial connectors installed in the faraday cage - or better, a coupling barel inserted through the wall. This should cost little in time and money to validate.
Next, select components for a diode-bias RF switching scheme workable at 2.4 GHz, and install this between the antennas, so that by changing the DC bias voltage you can electrically connect or disconnect the antennas.
Optionally, add a filter to pass only the frequencies of interest.
There will of course be loss is such a setup, but by keeping the overall distances short it should be workable. Applying gain would be very difficult, as in practical terms you would need to figure out which device was transmitting and enable only that amplifier, which requires a realtime understanding of the protocol.