More than half of the students I worked with demonstrated that they read much better than the single test suggested and there was another occasion where the difference between a student's actual performance and the test score was more than 7 years. I have suggested to the teachers I worked with that sitting alongside a child and hearing him or her read aloud from a text they actually are enjoying and then discussing the text together yields the potential for significant insights. It costs nothing additional to do this, can help to build community, and because it involves choice and performance, allows the teacher to build a more complex understanding of the learner, while affording the learner an occasion to deepen his/her metacognitive understanding.
The other potential problem with reliance on a computer generated 'reading level' is that some teachers and learners begin to doubt their own insights, especially if the test score is overly privileged. Whereas it's wonderful to help students make appropriate text choices and a test score can in some cases be helpful, we should not lose sight that lexile levels at best hint at performance. Prior knowledge, interest, confidence, and choice greatly influence how one reads.
My advice: sit alongside and come to know.
I'm sympathetic to the idea that at various points we're not challenging readers enough, but I'm deeply suspicious of the idea that a pervasive emphasis on quantitative measures of text complexity has much value.