'Inconsistent fundamental metrics (D/E, ROE) across agents in the same run; Valuation Analyst results dropped from final output; empty news sentiment data'.

Title

Inconsistent fundamental metrics (D/E, ROE) across agents in the same run; Valuation Analyst results dropped from final output; empty news sentiment data

Environment

• OS: Windows
• Python: 3.14
• Poetry: 2.4.1
• Repo commit: pulled 2026-08-01, range 09dd331..6c41ae8, project version 2026.7.10 (poetry install output)
• LLM provider/model: Anthropic, claude-fable-5
• Data source: FINANCIAL_DATASETS_API_KEY (paid tier)

Steps to Reproduce

1. poetry install
2. poetry run python src/main.py --ticker AAPL
3. Select analysts: Bill Ackman, Michael Burry, Nassim Taleb, Stanley Druckenmiller, Warren Buffett, Technical Analyst, Fundamentals Analyst, Growth Analyst, News Sentiment Analyst, Sentiment Analyst, Valuation Analyst (11 total)
4. Select LLM model: claude-fable-5

Issue 1: Inconsistent D/E ratio across agents in the same run

• Fundamentals Analyst reported D/E: 0.89
• Michael Burry reported D/E: 0.89
• Stanley Druckenmiller reported D/E: 1.34
• Bill Ackman reported D/E "at or above 1.0 across many periods"

Same ticker, same run, same underlying data source — expected these to match or at least be clearly derived from consistent inputs.

Issue 2: Inconsistent ROE across agents in the same run

• Fundamentals Analyst reported ROE: 119.91%
• Bill Ackman reported ROE: 151.9%

Issue 3: Valuation Analyst output missing from final table

• Console log shows ✓ Valuation Analyst [AAPL] Done
• Valuation Analyst does not appear in the final "AGENT ANALYSIS" table
• Portfolio Summary bullish/bearish/neutral counts sum to 10, not 11 (the number of analysts selected), confirming one agent's result was dropped before the final tally

Issue 4: Empty news sentiment data for a heavily-covered ticker

• News Sentiment Analyst returned "total_articles": 0 for AAPL
• Sentiment Analyst also showed 0 weight contribution from news, relying entirely on insider trading data
• AAPL is one of the most heavily covered stocks in the market; 0 articles suggests a broken or empty feed rather than a genuine absence of news

Issue 5: Warren Buffett agent appears not to use the same data as other agents

• Buffett's output stated "no market cap or margin of safety data—can't judge price vs value"
• In the same run, Fundamentals Analyst reported P/E: 34.98, P/B: 41.94, P/S: 9.66 — suggesting valuation data was available but not passed to (or used by) the Buffett agent

Expected Behavior

All agents drawing on the same underlying fundamentals data for a given ticker/run should report consistent values for shared metrics (D/E, ROE, valuation ratios), every selected analyst's result should appear in the final output table and portfolio tally, and sentiment data should reflect actual available news coverage for widely-covered tickers.

Additional Notes

Happy to provide full console output/logs if useful for debugging.