1) It ain't "Skiba's FTP test" - the critical power paradigm has been around for several decades, and I'm the one who originally suggested using it to estimate functional threshold power based based on a 3 and a 20 min effort (see this post from 2002 where I brought the critical power concept to people's attention: http://groups.google.com/...g/dbaf2ed122589566).

2) The reason that you don't want to add in the intercept is because the relationship between work and time isn't really linear (as assumed by the calculation). Ignoring the intercept gives you a slightly lower value, which helps correct for this non-linearity.

3) If your 3 min power increases and your 20 min power does not, then yes, you would calculate a reduction in critical power. That is, in fact, what would happen if the increase in 3 min power were due to an increase in anaerobic work capacity (which isn't quite the same as anaerobic capacity, at least as how I use the terms). OTOH, if the increase in 3 min power were due to an increase in fatigue resistance due to enhanced aerobic energy production, then your 20 min power should go up as well, and hence the slope of the line would increase.