But at least the 50/50 is within the bounds of reason, which is a probability bounded by zero and 1.
You can, in practice, have a defect rate of >100% in case the replacement(s) also fail, something I personally experienced with a certain harddrive vendor.
Any % numbers thrown around here are being estimates based on a data set that is massively biased toward the cases having a problem with the product (as I tried to illustrate through my completely ... statistic that I solely based on my
one defective device), they're not only useless but IMHO contraproductive.
My suggestion would be, instead of continuing an endless speculation in an echo-chamber, to collect hard (and proveable) facts as an effective base for any further discussion with the supplier in question. Which, in my experience, works better toward archiving a positive reaction (or change of behaviour) from a company as compared to an unsubstantiated approach.