Allow me to take a stab at answering this question.
While the "fact" of evolution is a completely observable phenomenon with surprisingly many direct supporting evidence, the evidence for the "theory" of evolution are far from direct. The theory is at best a conjecture from a set of evidence and data that directly point to what many creationists refer to as microevolution.
As a supporter of evolution on this board once mentioned not too long ago, not believing in "macro"evolution is like saying since we have only observed directly a second, a minute, an hour, etc. but we've never directly observe a thousand years, a thousand years must not exist at all. And this is the extent of how supporters of evolution think what creationists believe. In fact, the creationist sees a problem more complicated than combining a shitload of mini-whatevers to get a super-duper-whatever. Time is linear. Random mutations and natural selection are not.
Random mutation is self-explanatory. Over long periods of time, we may observe many of these random mutations in a population. While most of these mutations probably have no weigh in the "fitness" of the organisms to survive or reproduce better than others without the mutations, natural selection, a prejudicial process, is suppose to weed out the mutations that give negative weight to the survivability of the organisms and reward the ones that give positive weight by allowing them to be passed onto the progenies of the carriers. Natural selection may be guided by something such as some kind of natural selective pressure or sexual selection. These guidances change from place to place and time to time. For example, a mutation to give an organism thicker fur might be beneficial in Alaska and be deletitious in the equatorial region. Everytime the allele frequency of a population changes due to either a single beneficial mutation that have sufficiently been passed to enough individuals or accumulation of enough mini-beneficial mutations to give a noticable change in a population we call "micro" evolution.
The problem comes in when we start to assume that given enough time (millions of years) and enough of these mutations (many of which are neither beneficial nor deletitious) a population could be changed enough to have some kind of reproductive barrier to the parent population. In fact, we have never observe of such an instance when a population is isolated long enough from the parent population that a new species is resulted from the isolated population.
Yes, we have observed many species that are very physically and genetically close to surrounding populations, such as the different populations of birds observed by Darwin while traveling on the Beagle. And because of such observations, many have conjectured that these populations of birds must have resulted from a common ancestor because of their similarities. The question that comes natural for the creationist is how do we know these birds weren't already there?
Perhaps the problem lies in the common person's inability to comprehend such long periods of time and the rediculous number of generations and evolutionary steps taken for (say) mammals to emerge from reptiles. Maybe some people, with a bit of stretched imagination, can see how such changes can occur between one species of bird to another, let alone the different families or orders.
I guess what I'm trying to say in too many words is that the reason the theory is not scientific is because too much of it is conjectured rather than formulated from direct evidence.
™ Version 4.2