Indeed, some exchanges truncate their order book (BTC-E for example).
But the bid/ask sums may not be that meaningless, their scope is just limited to the provided range of orders.
Eventually, accuracy of this metric depends on the distribution of orders in the book.
Perhaps one needs some indicator of order depth that is less sensitive to truncation? E.g., weighted total sum of order size (BTC), where the weight decays exponentially with the difference between order price and current price?
With such a formula, one could compute relatively tight upper and lower bounds for the indicator, even from a truncated order book.
Its a good idea and one that I've thought of before, its not something I have the time or maybe even the skills for but Im sure there are people who could do this, in fact Im sure some probably already have.