Tyll's compensation target is a "diffuse field" look alike, he didn't implement harman's compensation on his default graphs (something was supposed to come up some day with a more advanced and maybe not free interface for graphs, where among other things you would be able to change the compensation applied. but nothing yet).
and about exporting the impulses, you do well to specify that the sample rate must be the same as what the playback system will use. that is indeed imperative for a proper result. so I would suggest that people do it for different rates(so they have them when they need them) or make sure they know the sample rate of whatever they'll use(on my phone using viper4android for convolution I have to run an impulse for 48khz). the one I use for foobar will take one impulse/wave file per channel. and I'm not sure that everything will accept 32bit wave files.
so that aspect of the process may require a case by case approach.
I just flew over the page so I can't claim I would have caught up a mistake if there was one, but it seemed pretty well organized.
for those who aren't so great with EQ done by ear, but find the result too cold, maybe suggest a way to tilt the final result a little. I believe many people will actually prefer that.
it would be nice also for people to have a mean to use the EQ values in an actual EQ when they don't have access to convolution but have a descent EQ at hand. the tricky part then is to select the type of EQ in REW that matched theirs. but I must confess that IDK how to do that without measuring anything. I guess just having the Q values the same way they do on their EQ would already let them get a close enough result.
IDK I'm really a noob at those stuff.