HOUSE RULE
I propose the following very simple house rule for Unreliable weapons (I'll give feedback as we advance in our campaign):
1. A Chaos Star rolled while shooting an Unreliable weapon starts a Misfire sequence (this is independent of the actual result of the shot).
2. Take as many difficulty dice as the Unreliable attribute of the weapon and roll them.
2a. If all of them give a Chaos Star as a result, then a major Misfire happens: the weapon explodes and causes 1 Critical Wound to its user (draw until you find an appropriate one). Roll again the dice and cause another Critical Wound to the wielder if all results are again Chaos Stars... continue with this process until not all results are Chaos Stars.
2b. If, in the first roll, not all results are Chaos Stars, then a minor Misfire happens: the weapon is jammed and its wielder must spend one Manoeuvre to solve the jam before being able to shoot again... alternatively, you can interpret the number of Failures and Banes obtained as a measure of how badly jammed the weapon is.
REASON WHY MY GROUP DOES NOT LIKE THE STANDARD UNRELIABILITY RULES
My group is not too happy with the Unreliable weapons rules for the following reasons:
1. We think the penalty for a misfire is too low (only 1-2 normal wounds).
2. We think that the probability of a misfire is too high (1/8 for Unreliable 1; 1/64 for Unreliable 2 if rolling two difficulty dice). Also, using an Unreliable 2 weapon in normal circumstances would never result in a Major Misfire (does not seem too realistic for Gunpowder weapons).
3. The rules as written are not scalable, meaning they could not be applied to Unreliable 3 or Unreliable 4 without giving a ridiculously small chance of a Misfire happening.
4. Historically, the effect of a misfire could be devastating (having a pistol explode in your hand is not too healthy), but the possibility of misfire was low (a group of 20 soldiers shooting arquebuses did not lose 1 soldier to misfire every 3 shots as a 1/64 possibility suggests).
ADVANTAGES OF OUR HOUSE RULE
With this rule, major Misfires can be simulated but their probability is low enough:
1. For experimental weapons (i.e., Unreliable 1), the chance of a major misfire will, in regular circumstances, starts at 1/64. This seems appropriate considering that only crazy people (example, Skaven) are reckless enough to use them.
2. For regular Gunpowder weapons (i.e., Unreliable 2), the chance of a major misfire will, in regular conditions, starts at 1/256. This seems to be reasonable... a "2nd rank mishap" (i.e., getting 2 Critical Wounds) would have a 1/256*1/64 = 1/16384 chance of happening.
3. You can simulate even safer Gunpowder weapons (i.e., with Unreliable 3, etc.). This ways, you can adjust the probabilities to what you think should be reasonable.
4. You can achieve very very bad Misfires (very Warhammerish and historically realistic) in which the wielder of the weapon suffers very critical wounds or even dies, but the possibility is very low (I can imagine such a roll would be very fun indeed)... If you decide to use this house rule, I propose adding to this post any casualties caused by the rule (a kind of Hall of Misfires Fame).
PS: My excuses for any misspelling or grammar error; I'm native Spanish speaker (actually from the Old World Estalia).