Thanks for the actual measurement. Did you put that diameter in the tool database for that bit?
BTW, my preference is to check runout at the tip, not at the shank. On my Makita it would turn a small diameter two flute bit into almost a single flute real quick. That obviously gives you a very poor effect on cut finish and plays heck with speeds & feeds.
…or maybe someone’s tried the DA30-0250 shown here? (“Collets 100% compatible with Kennametal Erickson collets”. What the heck is that, and does the Makita RT0107C indeed have that–versus the ER collet chucks?)
The “ER” collet system, developed and patented by Rego-Fix in 1973, and standardized as DIN 6499, is the most widely used clamping system in the world and today available from many producers worldwide. The standard series are: ER-8, ER-11, ER-16, ER-20, ER-25, ER-32, ER-40, and ER-50. The “ER” name came from an existing “E” collet (which were a letter series of names) which Rego-Fix modified and appended “R” for “Rego-Fix”. The series number is the opening diameter of the tapered receptacle, in millimetres. ER collets collapse to hold parts up to 1 mm smaller than the nominal collet internal size in most of the series (up to 2 mm smaller in ER-50, and 0.5 mm in smaller sizes) and are available in 1 mm or 0.5 mm steps. Thus a given collet holds any diameter ranging from its nominal size to its 1-mm-smaller collapsed size, and a full set of ER collets in nominal 1 mm steps fits any possible cylindrical diameter within the capacity of the series. With an ER fixture chuck, ER collets may also serve as workholding fixtures for small parts, in addition to their usual application as toolholders with spindle chucks. Although a metric standard, ER collets with internal inch sizes are widely available for convenient use of imperial sized tooling. The spring geometry of the ER collet is well-suited only to cylindrical parts, and not typically applied to square or hexagonal forms like 5C collets.
In addition to what the others said about the trim router’s poor performance, it’s not only the runout on the collet with no load that matters, but the deviation under load. And this does not only concern the router axis/collet stability, but also the deviation of the entire holder/Z assembly. The router and the way it is fixed to the machine is not ideal. The router is slid very far downwards into the 65 mm holder, prolonging the distance between actual cutting edge and the fixation point on Z slider and finally on X gantry. There are strong leverage forces that act on the system at the moment when the machine drives the cutter trough the workpiece. You might be interested in the discussion from the other day (@Marty 's wooden top and bottom box fit), where I tried to explained it.
The conclusion is that you can expect very much improvement by switching to a spindle, not only because of higher quality of bearings, axis and collets, but also because you don’t have to slide it that much downwards into the holder (see here vs. here). This way you keep the distance between the cutting edge of the milling bit and the fixation of the motor on the Z assembly shorter, thereby reducing leverage forces exerted during mechanical load.
Furthermore, if you use a spindle, due to the type of motor a spindle is, it will not be slowed down by the mechanical load during the feed, because frequency-controlledinduction motors work in this way. The hand trim router, on the other hand, is a universal motor (= a commutated series-wound motor) where, due to its characteristics, the speed depends on the mechanical load, i.e. it is slowed down by the load (what a more or less sophisticated electronics tries to prevent) and has only one speed(load)/torque combination in its curve at which the motor attains a balance between speed and torque, whereas a spindle has a constant torque over its entire rated speed range (e.g. constant 0.9 Nm from 6000 to 24000 rpm for a 2.2 kW spindle).
– Source: Traute Meyer, CC BY-SA 3.0, via Wikimedia Commons (rotated and mirorred to reflect axes of Image 2 and with comments added by Aiph5u)