First off, OF COURSE discriminatory screening will induce those actually leaning towards terrorism to make themselves more invisible (i.e, to gain a position in the relatively unsupervised group rather than the more supervised group).
This is just a perfectly normal arms race, and there's nothing wrong with that.
HOWEVER, any such added effort the would-be-terrorist must make in order not to get busted is a COST for him, one way or the other.
By always following a "one-step-behind" policy (it is impossible for the government to be one or more steps ahead) towards these perpetrators, the perps will enter the diminishing-returns-zone, where the increased costliness necessary to remain effective will become a barrier to their plans to begin with.
I.e, the number of attempts that will be successful, ALONG with the number of attempts tried will plummet/be significantly reduced, until we reach what we could call the "acceptable risk"-zone.
Secondly, let us say that the average time for a full body scan of any passenger is half a minute.
For a domestic flight with 300 passengers, at least 150 minutes, i.e, 2 and a half HOURS of screening time will be needed to scan everyone. If 20 minutes of total security delay is acceptable, this will require the installation of about 8 or 9 scanning devices and a similar number of separate operating crews on every airport that is an initial take-off site for such planes.
Of course, this is WAY too expensive, and at the very least, something that cannot be installed overnight, or even within a year or so.
What is, therefore, the necessary result?
SAMPLING PROCEDURES of passengers to be scanned WILL develop, whether we like it or not!
Since, in the given example, a 10% sampling pool will suffice for a single scanning device in order to be within the 20 minute delay, the only rational thing to do is to pick that 10% on basis of characteristics well-known to have within their midst a gross over-representation of terrorists, rather than picking our sampling pool out of a confused, irreflective policy, which is precisely what we would get if you, for example, left the criteria for sampling up to thousands of half-educated airport personnel.
That terrorists then will gradually shift their characteristics away from the initial sampling criteria should be expected but should merely result in a requirement of continuous monitoring, and a willingness to change the sampling criteria as the situation evolves.