Not using it, but AFAIK it uses a bloom-filter internally which is a datastructure that is known to come with unavoidable false positives. AKA: Even if the database was 100% free of errors and only contained actually transphobic people, the extension would still mark a lot of completely innocent accounts as transphobic, simply because the data-structure returns wrong results at times.

Based on that alone you should treat any and all results as “warrants a closer look” at most. With even that being a questionable use: People with dark skin are more commonly non-EU foreigners in Europe than white people, but every sane person agrees that the police checking dark skinned people more often is racial profiling, which is racist, should be illegal, and is just a bad thing to do. So taking a closer look at people because of something they cannot control that isn’t directly tied to them personally (skin-color or positive match in a bloom filter) is a questionable practice to begin with.

Lastly: What does it change? The validity of an argument is not affected by whether the person who made it is a transphobe or not.