For instance, if Yale had $1 billion in venture capital investments two decades ago—which is my educated guess of what it may have had, based on an overall private equity portfolio of $2.5 billion at the time—and just that portion of its fund had compounded annually at that rate, it would mean Yale’s VC portfolio alone would be worth $498 trillion today, or roughly 6.4 times the entire world’s GDP.

It’s not. Through June 30, 2015, Yale’s total endowment was valued at $25.57 billion. That’s up from $23.89 billion in the middle of 2014—during which time its exposure to venture capital has climbed from 13.7% to 16.3%—but still considerably less than half a quadrillion dollars.

So Fortune reached out to Yale and received an email response from chief investment officer David Swensen, who is credited with an asset allocation model that favors high exposure to “alternative” assets like venture capital.

He said that the 92.7% figure was a dollar-weighted internal rate of return (IRR). Were it to have been time-weighted, the figure would have dropped to 32.2%. And there was no compounding at the 92.7% rate, as “such a calculation assumes reinvestment of all proceeds from the portfolio during the period at the same rate of return for the remainder of the period.”

Swensen acknowledged that Yale uses time-weighting for assessing the performance of its marketable securities (e.g.., public stocks, bonds, etc.), because the endowment can choose when to buy and sell them. Sales within an illiquid asset class like venture capital, however, are primarily at the discretion of outside money managers.

From Swensen’s response:

The Internet boom comprises the beginning of the twenty-year period in question, during which Yale’s venture capital portfolio generated triple-digit single-year IRRs in three of its first five years, culminating in a 701.0% IRR in fiscal year 2000. Given the mathematics of the IRR calculation, a spectacular first five years of performance will have significant influence over the IRR of any longer term period. An anomalous period such as the Internet boom highlights this limitation of the use of IRRs, which is why we use several metrics to analyze the portfolio’s performance. In addition to the IRR, we consider net multiples of invested capital, dollar gains, and various other metrics. That said, we believe that dollar-weighted IRRs, though imperfect, are more appropriate than traditional time-weighted returns for the performance analysis of illiquid portfolios. Furthermore, utilizing dollar-weighted IRRs allows Yale to benchmark its performance in an apples-to-apples fashion using data services such those provided by Cambridge Associates, which generally report pooled IRRs.

Mystery solved.