Facebook said in a blog post on Friday that it has been miscalculating more of its viewership metrics for advertisers.

One miscalculation is a discrepancy between the number of likes and shares Facebook shows for web links through its Graph API for advertisers and through its mobile search field.

Facebook has also been miscalculating the number of likes and reaction emojis that page owners see for their live videos.

This is the third time Facebook has admitted to misleading advertisers since September, when it was revealed that the social network had for years been inflating a key video-viewing metric.

Neither of the errors Facebook revealed on Friday is as important for advertisers as the previously misreported numbers for video views and other products like Instant Articles.

But the discrepancy of the numbers of likes and shares for web links could affect the accuracy of recent and widely cited investigations by BuzzFeed and The New York Times into how fake news stories are shared on Facebook, according to Marketing Land’s Tim Peterson. Both investigations relied on Facebook’s data on likes and shares to show how fakes news stories can easily go viral on the social network.

Facebook said on Friday that it’s still “looking into” the discrepancy. The company recently acquired CrowdTangle, a tool used by media outlets to measure how stories are shared across Facebook and other social networks.

Visit Markets Insider for constantly updated market quotes for individual stocks, ETFs, indices, commodities and currencies traded around the world. Go Now!

By Laurie Sullivan.

Rakuten Marketing engineers believe they have uncovered a measurement flaw in Omniture, Google Analytics, Coremetrics and other analytics packages that measure the click-through rates (CTRs) and cost per clicks (CPCs) for Facebook mobile campaigns.

In Rakuten’s Facebook Measurement Divide report released Wednesday, containing the analysis of client performance data, the company reveals discrepancies between Facebook conversion tracking and Web analytics costing advertisers insight into 192% more attributable revenue and higher return on ad spend.

The cross-device campaigns analyzed reveal that attributable revenue only comprised on average 5.6% of the total revenue generated across mobile-only, desktop-only and cross-device campaigns — and as little as 2.4% for one retailer in the study.

Bob Buch, SVP of social at Rakuten — which supports attribution, affiliate, search, mobile, lead generation and more — said when the company began digging into clients’ campaigns it found that the CTR metrics were significantly higher and the CPCs quite a bit lower. “Omniture was missing 80% of the revenue, which explains why marketers are not investing more in mobile,” he said. “I’m not saying there’s something inherently wrong with the platform, but I do know it is not measuring mobile accurately for nearly every client we work with.”

Buch believes the tracking is inconsistent with what advertisers see in their Web analytics for several reasons. For starters, there are technological challenges that prevent conversion tracking on Facebook from functioning correctly, he said. In other words, there are additional conversions happening that are simply not recorded anywhere.

The report goes into more detail, outlining how post-click conversions are tracked differently by Facebook conversion tracking than in Web analytics. It also suggests that Web analytics platforms that rely on cookies cannot accurately track cross-device conversions because of the inherent challenges with identifying consumers across devices, and that some discrepancies are attributable to certain mobile operating systems and Internet browser combinations blocking third-party cookies.

Buch sees some of Rakuten’s bigger clients apply what he calls a “mobile multiplier.” He also says that it will be interesting to see what Adobe, Google and other platforms do to correct this discrepancy. When asked whether Rakuten sees this discrepancy with other social sites, he says, “I suspect this type of discrepancy would happen on any walled garden where there’s a mobile app linking to a mobile Web site, but truthfully the other social sites are not advanced enough to see the data at scale. We just don’t have the data.”