A 433MHz ASK transmitter/receiver pair should do very well at the ranges you are talking about. I have used them quite successfully in my Wicked Node and Wicked Receiver products, at hundreds of feet, so you should easily be able to use them to make it work over such a small distance. The transmitter can be quite low power because you can basically draws no current when not transmitting. The receiver on the other hand, needs to always be on and continuously outputs "data" that you have to sift through to extract actual transmissions from the noise. You could use the same exact product set to do what you want here but it will be a little bit longer than 1 inch, but you might be able to hack it into a 1 cubic inch with some creativity.
The modules you are using are only 0dBm (1mW) which is quite low power. The RFM12s that you are used to are normally 8dBm (6mW) and have a ready made, robust protocol. They are surface mount as well, so I'll assume you are using them mounted to a PCB.
My first thoughts would be to isolate RF issues from protocol issues. Set up a transmitter doing 101010101010 at whatever rate is required to transmit the data in your application. You seem to have a rate of about 2400Hz which is good for long distance without being so slow as to confuse any AGC in the receiver. Either use a scope on the receiver or set up something to detect this preamble like pattern, and see how far you can go with your current setup. This should make it possible to work out if it is an RF issue or a protocol one.
I don't know the HRR30 receiver, but most of the basic AM receivers have a non adjustable AGC in them. This means that it can be hard to work out exactly how long a preamble you need to bias the gain correctly, and also hard to work out what data rates are supported. Too short a preamble and the receiver will still be gained up and responding to noise. Too slow data rate and the gain will be all over the place. Your setup sounds fine, but may be worth investigating.
Bread boarding these transmitters and receivers is fraught with problems. I don't think that you need to resort to anything more than a quarter wave dipole though. Get the module onto a simple PCB and you will likely see a big improvement.
Another big assumption, but with projects like this it is often not worth implementing error correction. Just transmit several times and assume you will miss some packets. If that isn't acceptable, a send /ack system will be better then error correction.
Best Answer
Chipcon/TI CC110L (transceiver), CC113L/CC115L (transmitter/receiver).
Flexible sub-1GHz radio.
Claimed to cost $1 for pair of chips in volume.
A cheap devkit is available.