[Calvin Locke]

... four-way statistical tie

I'm gonna guess that they mixed and matched different ranges within the domain to get that result.

That is, they chose a large std dev to fit Yeb! in so they could report it that he's [statistically] tied for second place, and diminish the current also runnings-ahead of him.

[Robert A Cook PE]

They HAVE TO show Jeb!'s "improvement" to generate the improvement to get the enthusiasm to show the improvement - Jeb! MUST get above 10% to get any delegates at all.