Not None narrowing doesn't always work for nested field with list comprehensions

Bug Report

When a list comprehension is used to filter class instances with some property not None and another condition, mypy doesn't later infer that instances from that list has a not None value for the property.

To Reproduce

1. Write this code:

@dataclass
class HasOptionalField:
    is_default: bool
    field: str | None


def first_not_none_field(models: list[HasOptionalField]) -> str:
    models_without_none = [model for model in models if model.is_default and model.field is not None]
    return models_without_none[0].field

2. Run mypy on above code with below settings

Expected Behavior

Mypy should pass without errors

Actual Behavior

Mypy has error: Incompatible return value type (got "Optional[str]", expected "str") [return-value]
When removing the model.is_default condition and keeping just model.field is not None then mypy doesn't emit errors.

Your Environment

• Mypy version used: 0.971
• Mypy command-line flags: N/A
• Mypy configuration options from mypy.ini (and other config files):

follow_imports = "normal"
ignore_errors = false
implicit_reexport = false
warn_redundant_casts = true
warn_unused_ignores = true
disallow_any_generics = true
disallow_untyped_defs = true
check_untyped_defs = true
allow_redefinition = false
local_partial_types = true
strict_optional = true
strict_equality = true
warn_unused_configs = true
warn_unreachable = true
warn_no_return = true
no_implicit_optional = true
strict = true

• Python version used: 3.10.4
• Operating system and version: Windows 11

Use the provided dataclass and list-comprehension example as the first reproduction, then trace how mypy narrows attributes when multiple conditions are present. Add a regression test for this case and confirm that mypy accepts the return of the filtered list element without errors.