As per usual, science reporting fails to use precise language. It can be interpreted either way, although I think your interpretation is the slightly larger leap based on phrasing. In any case, it is far below the 70% (and not directed broadly at all scientific research) that GP states.

I went and read the article but have not tracked down the original paper.

It is at least 26%, because that was the percentage of studies that provided access to their data that proved faked or fatally flawed.

It may be substantially higher.