Trouble with dynamic schema markup on custom post types?

hey everyone,

we're running a pretty complex WordPress setup with a bunch of custom post types (CPTs) and custom fields, and using Yoast SEO (premium) for general SEO. all good there, mostly.

the main headache is getting dynamic schema markup to properly generate for these CPTs, especially when dealing with nested custom fields. the built-in schema options in Yoast are too generic, and manual JSON-LD injection for hundreds of posts is getting unwieldy.

we're seeing issues where the generated schema either lacks critical properties from our custom fields or throws validation errors because of incorrect type mapping. for example, trying to map a repeater field to an array of objects in Product schema, or correctly defining related entities from custom taxonomies. it's really messing with our rich snippet potential.

here's a simplified example of what we're encountering in the Google Rich Results Test (or similar structured data validator):

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Custom Widget v2.0",
  "description": "Our latest widget with advanced features.",
  "offers": {
    "@type": "Offer",
    "priceCurrency": "USD",
    "price": "149.99"
  },
  "review": { // This is a custom repeater field
    "@type": "Review",
    "reviewBody": "Great product!",
    "author": {
      "@type": "Person",
      "name": "John Doe"
    }
  },
  "relatedProducts": "widget-accessory-a, widget-accessory-b" // This should be an array of Product objects, not a string
}

--- Validation Errors ---
@type Product
  - Property 'relatedProducts' is not a known valid target for objects of type Product.
  - Unexpected type for 'review'. Expected 'Array' of 'Review', but got 'Object'.

what's the most robust and scalable way to programmatically generate or augment schema markup for highly customized WordPress CPTs without entirely ditching our current SEO plugin? are there specific hooks, filters, or perhaps lesser-known plugin extensions that handle this level of complexity when building custom schema types?

anyone faced this before?