Last week's explosive Quinnipiac University poll - showing Carl McCall surging far ahead of Andrew Cuomo - inflated the size of McCall's support with a questionable and undisclosed technique, The Post has found.

Several independent polling experts took issue with the Quinnipiac technique and questioned the value of the poll. 

The Hamden, Conn.-based Quinnipiac has a strong track record of accurate polling in past statewide contests, and its surveys are given great weight by the media and political operatives. 

The Quinnipiac survey, which inflicted heavy damage on Cuomo's Democratic gubernatorial campaign, claimed McCall had a "solid" 16-point lead among likely Democratic primary voters. 

That was a stunning turnaround from a Quinnipiac poll the previous month that found Cuomo leading by 15 points. 

But an examination of Quinnipiac's polling data from last week shows the number of New Yorkers backing McCall in the "likely" to vote subgroup was actually higher than the number of Democrats supporting McCall in the larger overall sample of "all Democrats" - a mathematical impossibility. 

The poll found McCall with the support of 34 percent of the 624 Democrats in the overall sample, or 212 voters. 

That overall group includes Democrats who may or may not vote in the Sept. 10 primary. 

Winnowing down that overall group to a smaller subgroup of only those likely to vote, Quinnipiac said McCall had the support of 47 percent of 482 likely voters. That translates to 226 voters for McCall - an increase of 14 voters compared with the overall larger sample, which is impossible. 

Quinnipiac polling director Maurice Carroll initially couldn't explain The Post's findings, conceding "the math is baffling, the number [of likely voters for McCall] should go down." 

But Carroll later insisted that the numbers resulted Quinnipiac's undisclosed decision to include Democrats "leaning" toward McCall among those described as solidly behind McCall's campaign. 

Carroll said the number of McCall voters increased in the "likely voter" category because of "a bunch of screening questions" that were then judged as to their significance, although Carroll said the same technique was not used in comparing Cuomo and McCall in the larger overall group. 

Quinnipiac's press release accompanying the poll never mentioned that "leaning" voters were included in the survey. 

Marist College pollster Dr. Lee Miringoff said he had never seen a poll "where the numbers move up as the sample gets smaller. 

"The [Quinnipiac] numbers don't add up, and there's a need for more explanation," said Miringoff.

For more news, entertainment and sports coverage, click here for NYPost.com.