Because it looks like what they want. The Marcott et al. "hockey" stick is spurious, A result of a cheesy method, not including data that was readily available and a few minor date "correcting" errors. The authors admit that the past couple of hundreds years are "not robust" but even highly respected institutions like NOAA include the "non-robust" portion as part of a comedic ad campaign.
The sparse number of data points in Mohtadi 2010 picks up the basic rough trend, but the Oppo 2009 indicates what is likely "normal" variability. When you leave out that "normal" variability then compare to "normal" instrumental variability, the instrumental data looks "unprecedented".
Since the lower resolution reconstructions have large, on the order of +/- 1 C of uncertainty and your cheesy method ignores the inherent uncertainty of the individual times series used, you end up with an illusion instead of a reconstruction.
The funny part is that normally intelligent folks will defend the cheesy method to the death instead looking at the limits of the method. The end result is once trusted institutions jumping on the group think bandwagon.
The data used by Marcott et al. is available on line in xls format for the curious and NOAA paleo has most of the data in text or xls formats so it is not that difficult to verify things fer yerself. Just do it!
My "just do it" effort so far has the tropical ocean temperatures looking like this.
The tropical oceans, which btw are the majority of the oceans, tend to follow the boring old precessional orbital cycle with few "excursions" related to other climate influencing events like Ice dams building/breaking, volcanoes spouting off and the occasional visit of asteroids wanting a new home. That reconstruction ends in 1960 with some "real" data and some last known values so there is not so much of a "non-robust" uptick at the end. It only includes "tropical" reconstructions and there are a few more that I might include as I find time and AD reconstructions to "finish" individual time series.