Did the guy doing the analysis at the shop blow off a bit of gas from the valve before doing the analysis? If so, and if he allowed the tanks to cool (or filled slowly in a water bath) then the analysis is probably ok.
It is possible to trap a (very small) amount of HP gas in the valve when it is shut off and the whip detached. If you hook the analyzer up directly (and immediately) you can get a bad number, because the gas in the valve is confined (and 3000 psi to 14.7psi allows a LOT of expansion of that "small" amount of gas!)
Also, as was pointed out, the sensors are heat sensitive (they're also PRESSURE sensitive - they're basically a small fuel cell)
In any event a 90' profile should be fine if you got a cut mix from 36%; you can go to 94' on 36% and not violate a PO2 of 1.4. I'd just stay out of the corner on the NDL, which is a good idea anyway.
BTW, if you want my opinion on analyzers, build one. There's a kit called the "YDI" one that I built, and it calibrated within 0.3% on pure O2 when I got done putting it together. That's better than the one in the LDS that I frequent (an expensive - commercial - Minox unit!)
Here's a link - $90 and an hour of your time will put you in business (plus whatever fittings you want to use for sampling)
http://www.oxygenanalyzer.com/YOUDOIT.htm