The Best Evidence

Research is a perennial topic whenever people talk about historical fiction (in fact it's easy to form the impression that it's the only topic, although it shouldn't be). I came across something today which prompted me to write this post, which is about the importance of thinking carefully about your sources.
Most of us take it for granted that primary source material is better than anything else and that secondary sources are unreliable. The way it was once put to me was: "Always go to your primary sources, because secondary sources are always wrong!"
Well, mostly...
I've been researching a new series to be set around the exploits of Robert Clive - "Clive of India". It's a fascinating but difficult business trying to switch from one period to another, but that's another story, or at least another post. Now, I knew there were slightly differing accounts of his escape from the French occupation of Madras (now named Chennai), but what I found today threw me completely.
I should explain that when the French seized the town in 1746 their commander, La Bourdonnais, negotiated very generous surrender terms with the settlement's Governor, in the course of which the British personnel gave their parole (ie, they promised not to escape or fight the French until they were exchanged for an equivalent number of French prisoners of war). La Bourdonnais however hadn't got the agreement of the French Governor-General in India, M. Dupleix, and the moment he took over the town he cancelled these terms and imposed much harsher ones of his own.
Now, all the accounts I've read so far agree that Clive, along with one or more of his colleagues, slipped out of Madras in November, after Dupleix's takeover, as they considered his actions made their paroles void. The main source for this seems to be a letter written many years later by Clive's widow (who of course wasn't in India and didn't even know Clive at the time). What I found today however was a letter by Clive himself, also written some years after the event, in which he says he made his escape at the beginning of October, when La Bouronnais was still in charge. A very small point but potentially important for my story.
Now, on the face of it, Clive's account ought to be pretty conclusive. After all, can you have better evidence for a person's movements than his own account of them? So once again the secondary sources must have got it wrong... Except for two things.
The first is that Clive is a notoriously unreliable witness. He wasn't given to telling outright lies, although he often exaggerated a bit, but there are instances where his memory for details simply let him down, and caused him to make mistakes even when the truth would have been more favourable to him (for example when giving evidence to Burgoyne's Select Committee in the 1770s).
The second is that it doesn't make sense. True, at least one officer, Ensign De Morgan, did escape from Madras before Dupleix took over - but that was in the chaos immediately following the French attack, and he had never surrendered in the first place. As Clive clearly didn't go then, he had no reason to do so later until the arrival of Dupleix's men at the very end of October, and every reason (under the laws of war as they were understood in those days) to stay put.
This is one case, therefore, where I think the primary source is actually wrong. It shows that all historical material has to be read very carefully, no matter where it comes from!