You can't add those numbers together - they are already cumulative totals of outstanding fails. (At least this is what FINRA tells the SEC it is). Thus on Aug 26th the total of all previously sold but still undelivered past the settlement date TSLA shares that FINRA admitted to was 720 shares.
I do find the FINRA daily information on short volume quite interesting. If you integrate it between NASDAQ reports of total short interest from 8/30 to 9/13 you find that 8,139,009 shares were sold short during that period but the short interest declined by 860,698 shares during the period. This means that 8,999,707 short shares were covered during the same period. However the total trading volume only had a grand total of 12,967,593 actual longs shares having been sold over this period (total of all sales NOT marked as short sales). Thus there could only have been a grand total of 3,967,886 shares actually purchased long and held during this same period which is a mere 7.7% of the trading volume during the period. The stock price soared from $229.15 to $246.96 as a result.
Welcome to the wacky world of greater than 50% of all sales being short sales. Needless to say one thing this means is that a large fraction of the buying from the short sellers comes from (theoretically previous) shorts covering. It makes me wonder about whether there is a game played (by non-market makers) to effectively have naked shorts by near-simultaneous shorting and covering (rolling forward the required delivery date). There isn't really a punishment for an FTD that doesn't last too long so once the "ball is in the air" they can keep putting off borrowing of the shares indefinitely by creating new shorts at the same time they cover the old ones such that the delivery is permanently pushed into the future 3 days at a time. Not something a retail investor could do, but hedge funds probably can if they're a member of the exchange.
Also worth noting that FINRA explicitly shows the SHO exempt short sales (the market maker naked shorts) too. During that period they were only responsible for 236,147 shares of naked short sales, or 0.88% of all short sales.
This all assumes you don't think FINRA is cooking the books. Personally I'm skeptical because of that magical period when there were no FTDs at all which lasted much, much longer than any other stock in history with over 1 million shares per day daily volume. I think that was just a mistake where their cooking methodology accidentally left a poo stain.