The technology actually isn't in question here; Illumina has entirely earned its preeminence in the field of DNA sequencing. That part of 23andme's process can be trivially duplicated in any sequencing lab, including the one a hundred feet thataway from where I'm sitting right now. Merely sequencing a genome and identifying SNPs is as close to a solved problem as anything in the explosively emerging field of bioinformatics.

What's in question is the validity of 23andme's analysis, and that's an entirely different matter. Bioinformatics is, again, an emerging science -- expanding at an amazing rate, to be sure, and full of enormous promise for the improvement of human health overall, but nonetheless still in its nascency. Treating bioinformatic analysis as an established part of medicine, to the extent of making "highly confident" (23andme's words, not mine) predictions of future medical concerns based on analysis of a relative handful of SNPs, strikes me as flagrantly optimistic at the very best, and potentially verging upon fraudulence at the worst.

That's the FDA's concern, and if anything it is only reinforced by the OP's experience. Were 23andme's analysis and predictions anywhere near the level of reliability the company claims on their behalf, it would not be necessary for one of their customers, having been terrified half out of his mind by the results of 23andme's analysis of his genome, to develop sufficient understanding of sequence analysis to replicate their process and identify their error.