I'd definitely recommend writing a script, even just a loose one, to cut down on the "um"s and the "so"s with long pauses around them. For instance, just writing a couple words for each topic or piece of evidence to keep you on track without sounding like you're reading from a script would help Or if you really don't want to script it at all, once you've stated the point slowly, restate the point more concisely and cut the first take out. You can even clap in front of the mic before you do each take so that there's a visual audio spike between each take, making it easier to find each one. I'd also recommend recording in different rooms of the house to test and see which one sounds best. Hope this helps, and keep at it