Well, 20 ft-lbs on a 6mm bolt into plastic is just asking for trouble. I cannot believe that is the correct torque spec. It would not be the first time that torque specs have been widely published and have been wrong. Same thing happens with the flywheel bolts on the '95. If you torque those little 8mm bolts to the specified torque, they simply break. 7.5 ft-lbs is all I have ever used with those little TB nuts and bolts on a plastic upper, and I have never had a leak yet.
[edit: here are some general torque specifications from the Engineer'* Handbook online for metric fasteners:
for cap screws:
for steel bolts:
All these give torque specs for DRY bolts. The same bolt tension or clamping force (which is really what we are trying to normalize when we torque fasteners) is achieved with less torque if the bolt is lubricated. According to this source, http://www.raskcycle.com/techtip/webdoc14.html
oiling the bolt reduces the torque requirement by up to 25%. With a dry lubricant, the torque requirement drops by half! When you analyze these and other similar data from a variety of sources, the conclusion is that 89 in-lb will provide all the clamping force these bolts were designed to provide.
willwren is correct to say that those bolts have a wide distributing nut set in the plastic that speads the clamping force widely around the base of the bolt. The problem is not that the bolts pull out, it is that the plastic can crack and distort when too much force is applied. This sealing surface is an area that is prone to distortion by warping anyway, and anything we can do to minimize warpage and distortion is a good idea.
I believe this is simply a case where some printed specifications are wrong. It is not common, but it does happen.]