Suji and related discard reading techniques rely on the assumption players are playing according to tiles' efficiency and are specifically going for ryanmen waits.
And they only tell if a tile is probably
safe or probably
dangerous. They are not supposed to be 100% accurate, but only go give an edge from randomly discarding anything.
Before using suji, you should first guess which kind of hand an opponent is aiming for:
• Discards a lot of middle tiles and no or few honours? -> Looks like chanta / junchan -> the player is most probably not following tiles' efficiency.
• Discards only from two suits? -> half/full flush -> TE will be less relevant.
• Calls pungs, discards tiles already discarded? -> all pungs -> use a whole different TE and won't have a ryanmen.
• Closed hand, discards tiles already discarded, doesn't sort his tiles? -> seven pairs -> also a whole different TE and won't have a ryanmen.
• Riichi early? -> the hand was almost ready from the beginning -> didn't had time for TE.
• Beginner -> may not know about TE.
In all those situations, suji and the like are of little use, if any.
The reason is because according to game record statistics from 東風莊's super high level tables, this so called discard reading, do not affect the risk of related tiles.
I don't know about those statistics, but if they're computed blindly without taking into consideration the probability a player is following TE, then they're not relevant.
Without knowing how they are computed and analysed, I can't tell if it is gold or dung.
Say for example you have a 49 aida yon ken. It may indicate that your opponent had a 4679 shape and kept the 67 only.
It may also indicate your opponent had only 49 and no other tile in the suit.
But without better tips, 58 are slightly more dangerous than other tiles (and 49 are 100% safe).
It's all what suji and the like can tell: what is slightly safer and what is slightly more dangerous. Only discarded tiles are 100% safe, and you can never tell which tiles are the waiting ones.