I wouldn't call it quite an "old wives tail", though in some case, a bit over stated. For one, it's mostly true, just as you describe. There IS often an enhanced error/uncertainty range at one or both ends. Generally the main (or only) problem will be errors that creep in as you first move off the rest state and all the initial inertial/static friction/slop is taken out. In the context of torque wrenches, the effect is more significant for clicker than beam or gage types. Once you're above that break-over threshold, it should be pretty even up to the top range. Unless, there is some component that is borderline and approaching the elastic/plastic yield, or getting into some friction/galling/binding situation, or dirt/contamination effect; then you might start seeing degradation on the top end too. Again, these upper-range error sources are more likely to affect cheap to mid range clickers. Theoretically, they could also affect dials, but they tend to be on the "high end" spectrum of tools (never seen a cheap import of those???) and so are likely unaffected with considerable built in "head room" to avoid this.
So, I think that "Always prefer to avoid using a gage of any kind near it's limits" is sound advice in general. Many times, especially for questionable quality imports, you're more likely to see problems as you approach those limits.
