It wouldn't be rewarding inaccuracy, as you put it. It would be penalizing an inaccurate initial solution...and allow the solver to have another go at it. This isn't some trivia thing where you take guesses and hope you're right, so no new information is really gained after you are told "This solution of yours is incorrect".
The method from the India semifinals -- we will check it for a minute and tell you during that time whether you're correct or not -- is significantly better (IMHO) than what is done at the USSC and UKSC. A time penalty would be applied, which is a reasonable middle-ground between indiscriminate guessing-and-declaring and the current system.
I wouldn't claim that this would necessarily be an improvement for the ACPT -- there maybe you just might not be able to decide between two possible letters for a given square or something -- but in these sorts of logic competitions I believe it would be.
The Art of Puzzles - Post a comment