I use a panheaded M3 screw with a matching nut for TO220.
Put the nut on top of the package so that any excess screw length sticks upwards.
It may be cheaper to buy machine screws (and nuts) of similar thickness at a nearby hardware store. My local screw dealer charges by weight, so I end up paying pennies (cents) for what I need in my hobby projects.
What matters most is that your screw fits in the mounting hole, which is about 3.5mm in diameter.
Total thermal resistance from junction to air is
Rja = Rjc + Rcs + Rsa.
Where: a = air, j = junction, c = case, s = (heat)sink
Rjc & Rsa are reasonably fixed by component choices.
Rcs is somewhat more variable - see below.
If your data sheet does not give junction to case thermal resistance (usually Rjc or similar)* then change manufacturers - this is one of the most fundamental thermal parameters and ALL manufacturers 'worth their salt' will supply it. You can look it up for a handful of TO220 packages to get a feel.
Your comment about varying with heatsink suggests you don't really understand the subject. Finding one of the many many good tutorials on the net would be advised.
Thermal resistances are like electrical resistance - they can be added in series.Heat flows from junction to case outer, from case to sink (cia washer and thermal paste, and from heatsink to air. So you get
Rthermal = Rjc + Rcs + Rsa
Trise jc = Rjc x Watts etc
Trise = Tjc + Tcs + Tsa
see prior paragraph for meanings.
Rjc is specified by the device manufacturer.
Rcs is set by thermal washer material, heatsink compound, pressure of mounting etc. It should be a minor contributor.
Rsa is set by heatink design and size, air flow etc.
Temperature rise = Rja x Watts.
Search eg www.digikey.com for TO220 and look at a few Tjc ratings.
This datasheeet
http://ww1.microchip.com/downloads/en/DeviceDoc/22057a.pdf
gives an unusually large number of Rjc and Rja ratings for various package versions of the same part and is a good starting point.
For the TO220 package Rjc is 2 C/W.
Rja here is from about 30 to 60 C/W but the heatsinks are "wimpy".
Large heatsinks with fans can be around 1 C/W - much less with much care.
A well heatsunk TO220 can manage Tja = 10C/W without too much effort.
But it can be far worse without due effort.
______________________
Edited:
The initial paragraph originally read as follows. It is evident that I wrote "case to sink" but intended "junction to case".
If your data sheet dies not give case to sink thermal resistance (usually Rjc or similar) then change manufacturers - this is one of the most fundamental thermal parameters and ALL manufacturers 'worth their salt' will supply it. You can look it up for a handful of TO220 packages to get a feel.
Best Answer
Lead resistance is not a problem. Look at the IRFB7545PBF data sheet. Lead cross-section is ~0.38 mm x 1.14 mm (minimum). Cross-section area is ~ 0.5 mm\$^2\$. This is equivalent to 20 ga wire, and has a resistance of ~10 mΩ/ft. A lead length of 0.3 inch will give a total resistance of 0.25 mΩ. Two leads, of course, will have a total resistance of 0.5 mΩ (less than 10% of the stated 5.9 mΩ). Total power dissipation for the package is 26 W (67 x 67 x .0059), which is well within limits for a TO220.
Under this load, each lead will dissipate 1 W. For a long run in free air, this is enough to melt the lead, but that is not how a TO220 is used. The leads will be soldered into a pc board with very heavy traces (to avoid delamination of the board), and these traces will serve to conduct heat away from the leads.
And for what it's worth, the data-sheet specifies the 67 A figure for a case temperature of 100°C.
ETA - According to a FAQ note at irf.com, the TO220 package has a current limit of 75 A, so the original suspicion that a TO220 can't handle high currents is almost correct.