You're just starting out so I'm going to take the liberty of giving you the full measure of my experience with the recorder.
There are, imo, a lot of oddities in the speech engine. However, I went to L1.14, and, for me, this didn't appear to be one of them. The only way I could get it to object to my pronunciation of mes, was to purposely booger up the pronunciation of devoir. The tendency to reflect some type of error in one part of the phrase, in other parts, or throughout the phrase, is an oddity of the speech engine. I have experienced this countless times - actually all the time.
It is possible that your microphone leaves something to be desired, but not necessarily. If you play back what you record and it sounds just like you, then it's obviously not your equipment. I have one laptop whose mic and speaker are crystal clear, and a “better” laptop that sounds like crap. So, I use a usb headset/mic with the latter, and that is crystal clear.
The speech engine will absolutely drive you up a wall if you let it. You simply have to resort to being your own judge sometimes, if for no other reason than it's tendency to confer false positives. It will very often give 100 although you've left words out or completely mispronounced something(s).
Sometimes it's objecting to very subtle differences. Sometimes it's just plain wrong. For example, many times I have only been able to satisfy it by eliminating an appropriate liaison that was present in the tutor audio. Sometimes you have to duplicate the rythym, flow and even the “lilt” to satisfy it, which you can attribute to the need for fluent speech. However, other times you can talk like you're drunk and it's fine with it. Occasionally, it will pinpoint your mistake, but as often as not, it will transmit something it doesn't like throughout the entire sentence, like dominos, which is obviously no help whatsoever.
Lastly, once you get to reciting longer sentences, you'll find that it very often will not give you sufficient time. You will simply have to speak very close to the same speed as the tutor audio. Many people complain that this doesn't give them the opportunity to practice the correct pronunciation, and this is absolutely true. If this happens to you, and you're obsessed with satisfying it, then you'll just have to say it as fast as you can. Chances are that will work, although not to your satisfaction. This has little or nothing to do with any “pause” on your part. The notion that it stops as a result of you “pausing” is easily debunked by simply saying the first couple of words of a sentence and then stopping. It invariably continues to record in total silence for a looong time, approximately equal to the length of the tutor audio.
It's a valuable tool and I can't imagine doing without it, but it is far from perfect. Do the best you can for your part, and otherwise, don't let it get to you.