I've had at least one report that the 2.6.36 imon module behaves as expected for another imon/mce user -- but it does still require v4l-utils ir-keytable. However, I don't think the problem here is actually in the imon code itself, but rather in the ir layer, so simply pulling in imon.c from 2.6.36 into 2.6.35.x isn't likely to be sufficient to fix the problem. I do have an updated ir layer backported for 2.6.35 that I've patched into the Fedora kernels, which does work as well, but I've already been told its too invasive a patchset for an Ubuntu SRU kernel.
However, another possible alternative... I *think* a ~10-line patch that adds back a modparam for choosing which ir protocol to use at module load time would work too. I actually put such a thing together for upstream, but in talking with others there, we decided against pushing it, as its expected that by the end of 2.6.37 development, we'll be uploading *all* keytables from userspace using udev anyway, so adding a modparam upstream would be irrelevant and/or redundant.