Okay, here is how the scores break down:
There are 4 quadrants (0-25, 26-50, 51-75, 76-100) for each category (academic, research, leadership). Based on how you stack up to the competition, you are placed in a quadrant and given a number in that quadrant that indicates your strength of weakness. That is how there becomes variability. Your scores per reviewer are then averaged (there are 2 reviewers). That score is then multiplied by the percentage net worth it is. All 3 sections are added and your total score is calculated. It is there that you are ranked in your entire committee.
If the 2 reviewers disagree heavily on your position in the quadrants, that is where your file is reviewed in February and the entire panel votes on where you rank. Each CGS/PGS application gets around 5 minutes of time. It is not as intense as you seem to believe.
If you disagree with any of this, feel free to contact NSERC - I did last summer when I thought my score was low (especially in academics). A 51 means you were above half of the population, but you are still far away from the top, and that can sink you. However, if the 2 people reviewing your file like your prof or believe you are in a good lab, your research component can get a great ranking even if you are not the strongest candidate.
So, yes, bias can easily influence this. And matters that are outside of your control (school, prof, tools), can become very powerful.
Best of luck.
Edit: That article does contain everything I wrote (except for the quadrants).