I'm not sure what your points are here, because you show cases where the RATs and MUL don't agree, and cases where there may be errors with both. My point is that sometimes units appear on RATs just to fill them out, because despite the thousands of units published, there are still gaps. (For example, the Chevalier Light Tank: maybe nobody but the SLDF really used that exact model, but the other states used local designs with the same stats.) Beyond that, mistakes also happen. Sometimes they're corrected through errata, and sometimes they're allowed to stay because their actual impact is minor.
If you feel you have found a mistake, the proper procedure is to report it as errata for the appropriate product. Do note that certain products are considered obsolete and will neither have errata threads nor be corrected. If any newer source disagrees with them, then the older product is usually considered to have been retconned unless the newer book is outright incorrect. TR2750, as one of the earliest Technical Readouts, certainly falls under this guideline. If you are uncertain what fix should be made, then before posting in the errata thread, post a question in either Ask the Writers or Ask the Lead Developers, but understand that it will take some time to get an answer, and that the answer you receive might not be what you wanted.
If you feel that the Master Unit List has an error, then this is the proper thread to post in. A unit's appearance on a RAT can be a point of evidence, but it might not be enough for inclusion on the MUL. On the other hand, a unit not appearing on a RAT is not a point of evidence in either direction. On the gripping hand, there may be internal discussion that places or removes a unit from the MUL without corroborating evidence in print.
BattleTech is a hugely detailed and sprawling universe, covering about six centuries "on-screen" over a couple hundred sourcebooks and novels. No one person can remember it all. Mistakes happen, and it is very much appreciated when the fanbase brings them to the developers' attention, so they can be fixed. All that's asked in return is patience on answers, and a willingness to accept that reports might not be acted upon or result in changes that the reporter might not personally agree with. "Hoarding" mistakes - that is, not reporting them because they're in your favor - or rejecting direct developer input is generally considered poor form, and is just going to prolong arguments.
Whoof. Sorry for the long post, and my apologies if I stepped on any toes, but I worked hard to develop the current errata system at CGL, and I know how hard everyone works behind the scenes, especially the MUL guys.