Yes, exit polling is subject to margins of error just as opinion polling is. But this is a "teachable moment" for understanding why the polls showing Trump performing worse against Hillary than the other Republican candidates should be taken with a healthy dose of skepticism. From the same exit poll and same pool of 1,363 respondents:
We get, according to the way the results were calibrated, adjusted, and extrapolated based on presumed voter profiles, within the same exit poll two separate breakdowns contradicting one another vis a vis the overall outcome.
Trump is shown leading by five points among men, who make an assumed slight majority of the Republican electorate. Conversely, Cruz is shown leading by five points among women, who comprise slightly fewer than half of the total. This suggests a Trump victory.
Yet when we look at income, it appears as though Cruz unequivocally won the state as a whole since he beat Trump by four points among those earning under $50k/year and by two points among those earning more than $50k/year.