No time for a 'hi, im new lol' thread. I have issues to discuss.
I've been working on a system for a few weeks now. At it's core, the system's challenge resolution mechanic is 'd100 roll high'. It's almost like d20 multiplied by 5, which means that mid to high levels frequently involves adding three-digit numbers.
Is this a flaw?
In a nut shell, yes.
The longer answer is, as others said, that its not a flaw its a cost, and the cost in math has to be worth the advantages in game play that it buys.
From my perspective, there is an important distinction to be made between one time costs paid at character creation - what I as a programmer call the 'compile time costs' - and those that occur repeatedly during proposition resolution - what I call the 'run time costs'. Because compile time costs occur infrequently, it's fine for a system to have a math intensive and expensive character creation process provided that you gain equivalent benefits in customability, granularity, and even reduced run time costs (by having many precalculated values). But expensive math during the run time is almost always a bad idea because math slows the human brain down so much (compared to a computer).
An example from gaming I'm fond of is the board game 'Risk'. Played by six human players, a game of 'Risk' can last 12-16 hours. Although this isn't obvious, Risk uses a very computationally expensive process to determine the outcome of battle, repeatedly rolling small quantities of dice and comparing them to each other to incrementally reduce the size of the forces. If you play 'Risk' on a computer with six human players making the decisions, the entire game can be completed in 45 minutes. This means of the 12 hours of play, 11 hours and 15 minutes are taken up by the mechanical action of dice rolling and the computational action of comparing numbers and less than 45 minutes of the game are taken up by decision making and real interaction between the players. Needless to say, I don't play Risk as a board game anymore.
There are things I like about d% systems, among them is the granualarity of 'leveling up' (see BRP/CoC), but a d% system better have very few modifiers to the rolls and certainly your end numbers or modifiers need to be small. My suspicion as a designer is that you've not been exposed to enough game systems to strike out on your own yet, and you are cribbing too much from D20. Two digit addition strikes me as plenty; why are you ever getting up to the levels of three digit addition? Get away from being trapped in D&D thinking. I think you'll find there are ways to make the system work without heavy reliance on run time modifiers. Those run time modifiers in D20 are their to provide granularity, but you need them less because you are providing your granularity on a different front.
Would you buy or play a game that expected you to add or subtract those numbers in your head?
No.
If you found yourself frequently reaching for a calculator [app], would you resent it?
Considering that I can easily add and subtract three digit numbers in my head, if I had to reach for a calculator to resolve an in game proposition (implying the math was even heavier than that), not only would I resent it, but I'd burn the book. The thing you have to consider is that its not the burden you put on the players in the game that really matters. The real constraint is how much burden you put on the GM. If the GM has to run your system for 4-6 (or 30!) opponents simulatenously and manage proposition resolution for the PC's, then it has to be pretty darn simple or it won't survive. If you have heavy math during proposition resolution, you better have a game that depends heavily on low drama for its primary game play and only occasionally dips into its very crunchy combat resolution system for dramatic scenes.