It's not necessarily about it being preferable as much as it is about knowing that it's possible to get that big a shift in alchemical values at all. I personally dislike that people near always look at % purity changes when that's not a linear increase, instead what they should look at is amount of alchemical value shifted. For rabidursus's experiment it seems to be in the range 2.20 - 2.30, distributed in each case in the following formats :
Increase/decrease/decrease/increase
inc/dec/dec/dec
inc/inc/dec/dec
So from his experiment alone you could conclude that either 1 or two elements can go up at once with the others going down.