We won't really know from the FINRA 'short selling' stats how much short interest reduced last week: it's very imprecise due to the way entire block sales of thousands of shares can be flagged by brokers as 'short' if even just a single sale there is from a known short seller.
As Ihor pointed it out numerous times the FINRA stats seriously over-estimate the amount of short selling - the true proportion of short sellers is closer to 10%-20.
I can confirm this based on analysis of actual transaction level logs of NASDAQ short selling activities: for example on 2018/09/18 there were 3,359,275 TSLA shares sold short, while the daily volume was 16,550,000. That's a true short selling percentage of only 20.3%, on a day that featured a major bear raid and a lot of short selling. (This data might be interesting to @Papafox too.)
FINRA stats also have trouble catching "unknown" short sellers (i.e. big short sellers with multiple accounts over brokerages and creatively moving around positions) and with 'dark pools' (interim trading platforms that effectively anonymize the source of the trades).
With all that in mind, this is the FINRA short selling stats for last week:
So yes, it was a lower percentage - but this is not necessarily indicative of true short covering. Friday price action definitely gave me the feeling of significant short covering.
I'm curious where you obtained the 3,359,275 shares sold short number for TSLA shares shorted on 9/18/2018. Last year I queried Dusaniwsky to see if he could give us a rough estimate of actual short-selling on a day when shorts were tagged with 50% of the selling. He felt he could not give an accurate estimate, with the batching of shares issue being a factor (as you mentioned). It seems to me that if you can get numbers like this, Dusaniwsky can too and once you have the actual number it's not hard to come up with an actual percentage of selling by shorts.
We've seen some pretty significant percentage of selling by shorts on days when we're told that short interest has not changed. Therefore, I assume that part of the percent of selling by shorts number is due to batching, part is normal market-maker short-selling as an expedient for trading, and some is part of algobots working the stock with both buying, selling, shorting, and covering to extract value and (I believe from watching the charts) for also manipulating the stock price. Of course some is also shorts opening new short positions in the stock.
Anyway, please let us know where that number came from so that we can better understand the implications.
Thx!