My approach for both question types was largely similar. It is important to spend adequate time upfront, not rushing through but also not dragging, while collecting all the relevant details. For MSR, when you begin answering the questions, try to return to the tab that contains the relevant information, cross-check it with your mental map, and decide whether you can confidently answer the question. If you are uncertain about multiple sub-questions, it is often better to skip them early on.
Time management is key. For example, if the MSR passage is dense or complex, the associated questions are likely to be easier, so it makes sense to spend an extra minute understanding the passage well. The reverse may also hold true. Pay close attention to the types of mistakes you are making. It is easy to overlook a key word in the tabs while skimming or to misread a question asking whether something can be evaluated, leading you to solve it unnecessarily and waste time.
For TPAs, it helps to use the answer choices early. Once you grasp the core of the argument, working backwards from the options is often effective. If you have read the argument and the options but still feel stuck, make your best guess and move on. The questions here are often rooted in core Quant or Verbal concepts. If you consistently draw a blank, it is a sign that reviewing those fundamentals would be worthwhile. Also, going through expert solutions afterward can help you catch insights you may be missing in the moment.
From your mock results, it seems that cutting down on errors in your stronger areas like DS could significantly boost your overall score. Additionally, it looks like you have been spending too little time on TPAs to get them right. If that is by design, that is fine, but if not, you will need to allocate a bit more time to solve them correctly. Try to build a strategy that lets you confidently solve the questions you know while also securing the easy and medium ones. You do not need to be perfect. Just play to your strengths and make smart decisions.