That is a valid point, I suppose it does depend on play style and how open and closed a GM or DM is, and how they respond to various events. However, you can tell certainly when it comes to, for example, the AC of a character; especially in a bounded accuracy system, it is fairly trivial to figure out the AC of an unknown and unfought monster thanks to the variability of dice; recording when you miss and recording when you hit gives you an easy way to create an AC range and narrow it down. This allows you to quickly qualify how much your modifier affects your chance to hit.
In addition, against saving against negative effects as your character, you will have a pretty quick understanding of whether you meet a DC or not depending on how constant the saves are for it.
A trait is more explicit, but it can also be quite a lot more niche and depending on the situation, much less impacting than a modifier.
In my opinion as well, the harder it is to quantify how much that 5% matters, the more uncertaintiy - and the more import - I would put on it. Personally, for me, I try to reduce risk as much as possible in an RPG (particularly in a potentially deadly game for my character) rather than necessarily explore more exciting features. Not knowing how much a 5% difference makes it a riskier prospect to, say, not take that 5% increase.