I believe the reason for this epistemological problem with political science lies in accounting.
Purely economic exchanges are somewhat easy to measure and assess with objectivity. It usually boils down to “somebody gave away A in exchange for B (and, of course, his counterpart gave away B in exchange for A)”
So the science of economics can readily identify many patterns on the data that can be collected through such observations. Say, for instance, the law of diminishing returns or the dynamical relationships between supply and demand. These and many other market patterns are percetible and explainable because we can observe and measure many of the variables involved in their occurence, like prices, costs, stocks, production and transaction volumes and so on. Of course many strategical variables remain concealed and introduce uncertainty, but still much of the action takes place on the open field, therefore allowing for a class of general forecasts which bear some similarities to the predictions of natural sciences (sometimes these similarites are greately overrated, though).
Political bargains are harder to quantify. They are most of the time implicit bargains that do not take place, because one of the “traders” doesn’t want to play chicken and get hurt. But nonetheless his behavior is shaped by the likely outcome he perceives from such a situation taking place.
The chief use of power is the (partial) canceling out of other, potentially threatening, sources of power. This cancelling might take place by domination or destruction, which would allow for objective measures with hard data, but most of the time it takes place by insinuation and demonstration and readiness and other forms of non-transactions hard to evaluate.
This kind of situation is very hard to measure since there is little direct data to be observed. Most of the action is not happening. Most of it consists of self-defeating profecies. Anticipation is the name of the game.
Of course the political strategies can be inferred from reveled decisions taken and not taken by each player, but such analysis cannot be carry on too far without getting too complex.
And you cannot generally engage in large scale political accounting because most of the consequential information is private.
And without objective accounting procedures it is difficult to agree on what’s really going on, much less on what are the scientific reasons and explanations for what we’re not even sure to be happening.
That’s why political science is a much more vague and “unscientific” than economics.