Are most 1/3 downward variances reasonable, while most larger ones aren't?

Two very interesting sentencing cases decided by the Eighth Circuit have me wondering this reasonableness rule of thumb may exist in the Eighth Circuit and elsewhere: most downward variances of 1/3 or less from the bottom of the applicable range are presumptively reasonable, while larger one presumptively are not.  Here are the unofficial summaries of the two Eighth Circuit decisions that have me thinking about this reasonableness rule of thumb:

• US v. Jones, No.  06-3616 (8th Cir. Nov. 2, 2007)(available here) [Benton, Author, with Bye and Shepherd, Circuit Judges]:  District court placed too much weight on defendant's overstated criminal history in granting a 54% variance and the sentence imposed was unreasonable; case remanded for resentencing.

• US v. White, No.  06-3781 (8th Cir. Nov. 2, 2007)(available here) [Bye, Author, with Riley and Benton, Circuit Judges]:  Evidence was sufficient to support convictions for receiving, possessing and distributing child pornography; no error in imposing an enhancement under Guidelines Sec. 3A1.1 while also imposing an enhancement under Sec. 2G2.2(b)(1) as the enhancements were based on separate victims; no error in imposing a two-level enhancement under Guidelines Sec. 2G2.2(b)(2)(E) for distribution of child pornography; on government's cross-appeal, the district court did not abuse its discretion in granting a 33% downward variance; sentence was not unreasonable in light of the mitigating circumstances the court identified.  Judge Riley, concurring in part and dissenting in part.

Of course, these two cases alone do not indicate that the Eighth Circuit or other circuits have a 1/3 variance reasonableness rule of thumb.  However, I have a sense that most downward variances of 1/3 or less are not even appealed and most get upheld absent a real problematic rationale.  And, conversely, I sense that many appealed downward variances greater than 1/3 are reversed absent a real compelling rationale.