a) how big are the butterflies you want to get?
b) do you want them to fill the screen, or have a nice background?
c) which camera?
I took my 7D + 70-300 non-L to a butterfly house recently, and my 12/20/36mm kenko tubes.http://www.canonrumors.com/forum/index.php/topic,307.0.html
for a sample (there's better ones i haven't uploaded yet). i started out with longer tubes, ended up with just the 12mm, and gave up on autofocus altogether by the end of it. (ok, so the 70-200L will beat it in autofocus speed, but like everyone's said, with tubes they just *hunt* all the time
). one finger on zoom ring, one of focus ring, one on cpl (damn non-L rotating element) and jumping around everywhere. i accidentally stepped on a lot of feet that day.
so, looking at the specs, the MM of the 70-200L 2.8 ii is 0.21x. on a 7d sensor, that means your framing width is 22.3x14.9mm / 0.21x = 10.6cm x 7.09cm, at max. mag.
on a 5D, you get 17.1cm x 11.4cm.
so either way, yeah, you'd need a big butterfly to fully fill those images. cf the image i linked above, that's uncropped, and the butterfly is about 4-5cm across, the whole image is maybe 7x5cm. so you can get the same scaling on a 7D with your lens. with a 5d you'll get a bit more space on the edges.
adding the +12mm ring gets you .28x or +25mm gives you .36x.
firstly, my advice is get the kenko tubes. 3 rings gives you +12/20/32/36/48/56/68 combinations. and all for the price of less than one canon ring (aside, are the canon rings sealed? that'd be the only reason i'd pay a premium for a bit of air-spacing)
secondly, my advice is to skip the rings, if you want the 70-200 2.8 for other things, get it and crop. if you only want to take insects, get the 100LIS macro. (with the money saved, get a teleconverter too?)
thirdly, it is possible to get used to MF using rings. your zoom ring also acts as a sort of focus ring once you put the tubes on, every focal length will only have a small focussing range, it can be easier to zoom in/out than move forward/back.
lastly (if you don't already do this). High Speed Continuous is your friend. crappy shots can be deleted later, memory is cheap, and you can always crop something good out of a wider shot. if you spend too long framing and getting closer to fill the frame, it'll just fly away eventually and you get nothing...