One inconsistency that we've run into trying to track selling and repurchasing by shorts is that the high percentage of trades involving shorts, often over 50%, is too high when you look at Fidelity and use it as a measure for short drawdown and covering. For example, in a typical day we see less than 200,000 shares drawn down for shorting at Fidelity and a smaller (often much smaller) number of shares covered.
I've been assuming that Fidelity represents nearly half of shares available to short, but maybe that's where the trouble lies. In a day with volume at 5M shares, there are 5 million transactions and if short selling/buying is at 60% of total, that means shorts were involved in 3 million of those trades. If the Fidelity drawdown plus covering < 150,000 shares, though, you would conclude that Fidelity's involvement in making shares available to short or cover is only about 5% of what's going on in the market, which seems way too low.
The question is how do you explain 60% of transactions involving shorts when the drawdown/covering at a brokerage such as Fidelity is so low? I have followed TSLA shorting extensively for the past year and this is the question that most bugs me.
One possibility is that market makers are doing lots of micro transactions, positive and negative, and that their volume dwarfs what a brokerage such as Fidelity is making available to the public. I just don't know enough about this side of the trading world to grasp the big picture.