When you start trying to measure +/- 1um, you need to ask at what temperature. Just because the instrument displays 4 or 5 digits to the right of the decimal point does not mean that the measurement is anywhere close to that. In the home shop or even on the factory floor, the best you can realistically measure is about +/- 0.0002'' ~0.005mm To get closer than that requires a metrology lab and closer than that is just not needed in most applications.
As you stated, parallelism of the anvals is what is the most important for accuracy.
I have a set of SPI digital mics that seem to work OK, and seem to match my cheap gauge blocks, but I'm not doing work for NASA. The most critical thing I normally do is getting shaft fits in bearings and have had no problems.