What you are saying makes total sense and seems reasonable but the reality is the books don't need the sharp action and don't really care about having a perfectly even balance. They will make more money just letting squares fire away at 1.9 or -110 odds and print money. Someone above said it perfectly, having the balanced action via sharp players will reduce variance but will eat into profits, and the books have the capital required to ride out any variance (entire seasons where squares win at a higher than expected rate for example).

edit: the books (at least the ones that open with respected lines, not the ones that copy the big books' lines) DO want sharp action when they are trying to set the right line. They will have a private group of handicappers that get to take shots at the new lines before they are publicly listed for smaller bet sizes, then they open the lines publicly for very small bet sizes. By the time the full bet limits are allowed and the square punter is picking a game 30 minutes before it begins, the books have no more fears.

on the topic of this paper, the arb'ing or picking off "off" lines is kind of a crowded space and opportunities don't last for long. I wouldn't get into it with high expectations of profit.

edit 2: source: i've been heavily involved in the industry on the betting / modeling side for years in the past.