First point, the 68.05% number is sloppy, if not completely specious. The author notes that there were "15,676 individual events" recorded in the tabulation log, but doesn't specify what an event is in this context. I assume that an 'event' is any pertinent action performed by a user or the machine and recorded in the log, but regardless an 'event' is clearly not the same a 'ballot'. Comparing an error rate derived from an ongoing stream of 'events' to the Federal 1:250,000 guideline meant for final ballot tallies is fairly close to nonsense. For an apt analogy, this is a bit like having a rule that says we must drive straight home after work, and then having someone claim we broke the rule because we made all sorts of left and right turns as we were driving home. Knowingly or unknowingly, it's bait and switch: invoking one thing and producing something else entirely.
To the main point, please be careful with the word 'certain' here. There are people who are 'certain' of all sorts of crazed things: that the Earth is flat, that space-aliens are being held in Area 51, that Tupac and Kurt Cobain are still alive and sharing an apartment in Queens. This report is trying to argue that these 'event' discrepancies are a sign of intentional malfeasance through a convoluted process: that the machines were intentionally designed to produce a high rate of 'event' errors so that ballots could be reassigned to the 'adjudicated' category where they could be manipulated without a paper trail. What the numbers actually tell us is that something went slightly sideways in the early ballot counting. Between Nov 3 (election night) and Nov 5 5200 ballots for Trump were counted (probably a normal part of the time-consuming ballot counting process), and about 200 ballots that had been given to Biden were corrected and given to Trump. Then at some later date — probably after an audit — roughly 2000 ballots (most of them for Biden) were removed for one reason or another. The report asserts (probably justifiably) that the Michigan system was cloddish: they didn't do proper software update or security fixes, etc. But it takes that observation and leaps first to the contra-evidential conclusion that exploits did occur, and second to the wild-eyed conclusion that these vulnerabilities were expressly intended to allow exploits.
Without knowing what these 'events' refer to, all we can say for sure is that this county had an unfortunate number of corrections to tallies: something in the range of 15% (2.2k corrections out of 15.9k valid ballots). For all we know that might mean machine error, malicious attack, or some ballot-counter stuffing a stack of ballots into the scanner backwards. Nothing in this indicates that the final totals violated the Federal 1:250,000 guideline, and in fact such a large number of 'events' could easily be interpreted as a normal part of trying to meet that guideline in the final totals.
Possibility is not actuality, speculation is not fact, and while the absence of evidence may not be evidence of absence, this ongoing game of "Don't Show and Tell" has worn out its welcome.