The cause is the people who run it and put skates like the ones Chan does (at Worlds) as the ones that they award the highest scores to. And if those are their values, they are out of sync with what is either right or appropriate for the sport, and everyone, except for them and a few of the winning skater's willfully blind fans, knows it.
See, I don't think that the people who run the sport hold values that "performances with superior skating skills and difficult content and several falls should win over clean performances with good skating skills and difficult content" (or however you want to define it).
There is and always has been a belief that skating skills should be rewarded, and that falls should be penalized. I think everyone agrees with that. The question is how to balance them out, along with all the other positive and negative factors in the performances.
The current scoring system is designed to build in specific rewards for each different kind of skill and specific penalties for errors.
The penalty that's built in for falling is the same at all levels of competition. I think when the fall deduction was introduced ca. 2005, the designers were thinking that they had to guarantee some additional penalty for falls beyond the negative GOE for falls on elements. And they chose to use 1.0 per fall, which was equivalent to what was then the value of one more minus on a triple jump.
That choice meant that each fall was very costly at lower-scoring levels of competition, and not very costly at higher scoring levels.
When you get a skater who deservedly racks up points for difficult elements, most of them well performed, and for program components, than subtracting 1.0 per fall isn't going to make much difference in their placements.
At most levels of competition, if a skater is far enough ahead on other skills that points lost to a few mistakes plus 1.0 off for each fall don't drop them behind a lower-scoring skater who didn't fall, it's because the skater who stayed ahead was that much better.
That can also be true at the highest, most high-profile events as well. But in choosing top-level champions, as opposed to choosing who's good enough to move on from a qualifying competition, it's more emotionally satisfying to see cleaner performances win.
But even if the judges lower scores PCS for the better skater with falls from what they would have given for a clean program, and even if they raise PCS for a less superior skater with a cleaner performance further than they would have if everybody had skated clean, they can't force the cleaner skater to win. They don't have control over all the pieces in the system -- e.g., base values of the technical content completed, second-half bonuses, lead carried over from the short program -- when assigning GOEs and PCS in the long program. If they mark each element and each component the way they see it, they may still be surprised and perhaps disappointed in the final result produced.
This is a bigger problem at the highest levels because 1) when the PCS values are high at the top there's more room for variation between the best, good, average, and weak skaters within that event, and 2) high-value elements rack up a lot of points when completed well, and the very highest-value elements (difficult triples and quads) rack up points even when they're failed.
When that happens a couple of 1.0 deductions for falls make a negligible dent in the lead that the stronger skater can build up with all the other elements and components -- and short program.
At lower levels, on the other hand, where the elements are worth less and the PCS range is narrower, even one fall can be a killer because the next-best skater is more likely to be closer behind.
So if we want falls to be similarly costly at the highest levels, where the most people are watching and the most people care strongly, then the system needs to be redesigned to make the penalties larger at levels where the overall scores are highest.
I don't know whether the ISU technical committees or whoever else is responsible for designing the system is already working on this problem. They probably do realize there is a problem but they may not have yet conceptualized a good way to solve it. The solution they tried to add ca. 2005 by formalizing a fall deduction in place of the original recommendation to judges to lower the Performance/Execution score was some improvement, but they didn't foresee at that time some of the ways in which it has more recently proven to be inadequate.
I propose making the fall deduction a percentage of the total segment score. But even if I had the power to wish a new rule into effect immediately, I would want to run simulations and shadow scoring of many competitions at different levels to figure out the best percentages to use.
Meanwhile, all the judges can do is call each element and each component as they see it. Unless you want them to ignore all the other criteria and manipulate numbers to make sure, as much as they can, that the cleanest programs will win.
Based on the good-humored energy of your posts, I somehow doubt that (if that were true, I would expect something, at least occasionally, along the lines of "damn sniveling punks, get off of my server!")
And I fear I am developing a strong tendency to be a real Jason Brown uber-his stuff is just such fun. Not to mention Zhiganshina & Gaszi 
