If you look at the SCA3300 datasheet you'll see its nothing like that accurate. https://docs.rs-online.com/cd88/0900766b815d7c81.pdf
Its not clear what the repeatable precision is though, but that's not really relevant to measuring static angles unless you want to go through a complicated calibration sequence, and repeat this regularly - the 0.7% sensitivity error includes drift over lifetime for instance.
For the MPU9250 you may be confusing the dynamic accuracy (gyroscope) with static angle accuracy (accelerometer). The only mention of 0.05 I can find in the datasheet is the pin-spacing in inches.
There may be some confusion about accuracy and precision going on - if you make a measurement, accuracy is your primary concern. Precision without accuracy will help with repeatability, but you will be repeatable wrong... Calibration of a device can greatly increase accuracy, but temperature dependence will probably need to be characterized in the calibration for these devices, and long-term drift may invalidate the calibration regularly too.