I want further to expand my response to this.
It is not a "super small increase," and in fact it is INCREDIBLY significant. What it boils down to is a binomial probability distribution, where the number of events is the number of spell casts, and a success is defined as a proc. Because I suck at doing distributions by hand, I used this tool:
http://easycalculation.com/statistic...stribution.php
Now, let's say we cast 2 SF with a 5% chance to proc the trinket. Let's calculate the odds of getting at least one proc. Our n=1 and r=0 (we will subtract from 1 to get the odds of getting at least one proc) with a p=.05. We can see that the odds of getting at least one proc are about 10% with 2 chances.
However, if those were instead SB we would have to factor in crit chance. To do this we'll use the same distribution to calculate the odds of getting 2 crits in a row with SB to give us the same two chances we had in the first example. Using your own crit rating of 27.96% the odds of getting 2 crits in a row are 7.8%. Therefore, by casting 2 non-guaranteed crit spells there is only a 7.8% chance that you will have a 10% chance to get at least one proc, which is less than a 1% chance of getting at least one proc overall.
To simplify this even further, at any given moment in time the chance 1 SF would proc the trinket is simply whatever the proc chance happens to be at that moment, as it will be a guaranteed crit. The chance that any other spell will proc it at that moment in time is whatever the proc chance is at that moment, multiplied by your chance to crit.
So you see, by using a guaranteed crit spell instead of a non-guaranteed crit, there is a very significant increase in proc chance that has nothing to do with RPPM mechanics, and everything to do with statistical chance.