Ben, this really depends on what you consider "significantly inaccurate" and what type of ascents and descents you are making. If you ride primarily on flats, that a have a few small hills, then I would say the elevation gain/loss data might not mean much. If you are climbing mountains, you will get excellent data, I can assure you of this. MB (I believe) uses the tracking points that the FR's upload to it, and it has a logarithmic (or is it algorithmic?) formula to compensate for any extreme outlying points that are obviously errors. Let me give you an example of the accuracy of feet ascended and feet descended on some workouts where I start and stop in the same point.... (I couldn't copy and paste the data, and I'm not sure that you can go to my own digest page, so I printed the page and took a picture. The things I do for you Ben...)

You'll notice two or three of those workouts have large discrepencies between the elev. gain and loss, however, on those occasions I either forgot to turn the device on before I started, or I had gone too many days without re-charging it and it may have died (only happened twice, as they last about 15 hours and I'm usually more conscious of battery life) You'll also notice that on the other workouts, where I was more prudent, many of the elev. gain and loss numbers are within .04% - 1% of one another. That's plenty accurate for me, I mean, whether I climbed 7000' or 7080', does it really matter?