Student ratings actually negatively correlate with learning
I think maybe I'll trust the decades of research on math education rather than your intuitions. For example, there is research showing that many more students succeed in calculus when the math is taught in context, not abstractly and plug and chug. One set of studies showed that such a course increased calculus pass rates from 60% to 90% (see the Wright State engineering math course, and see the ENGAGE Engineering project). Another large set of studies shows that student learning can double and graduation rates triple when active, participatory learning techniques are used by the instructor instead of traditional lecture (The PCAST report - Engage to Excel cites it).
I think maybe I'll trust the decades of research on math education rather than your intuitions. For example, there is research showing that many more students succeed in calculus when the math is taught in context, not abstractly and plug and chug. One set of studies showed that such a course increased calculus pass rates from 60% to 90% (see the Wright State engineering math course, and see the ENGAGE Engineering project). Another large set of studies shows that student learning can double and graduation rates triple when active, participatory learning techniques are used by the instructor instead of traditional lecture (The PCAST report - Engage to Excel cites it).