Tesseract has a horrific API and decades of technical debt. Add to this, it's also very sensitive to fonts. So... yeah, OCR is 'solved', but that doesn't mean it's easy.